D4.1 Device Selection

806999 – RADAR-AD

Remote Assessment of Disease and Relapse: Alzheimer’s disease

WP4 - Development of a technology-enabled system to measure identified functional domains via smartphone, wearable and fixed home-based sensors

D4.1 Device Selection Report

Publishable Summary

Wearables and smart devices are increasingly integrated into our lives, flooding the retail market and literature research. Besides lifestyle applications, they could be utilized as digital biomarkers in elderly care, through activity and behavioral monitoring. In this deliverable, we selected the most appropriate choices of wearable and smart devices, available in literature and the market, considering the three Tiers of trials that will be carried out for the RADAR-AD project. Consequently, we first examined the available sources for all current wearables, smart home (ambient) devices and smartphone apps. Their most important criteria, pros and cons and specs are fully listed for each Tier and category. Then, we produced a shortlist table of suggestions, including devices and apps that match the project’s needs. In the end, all partners conversed through biweekly WP4 device selection track calls, biweekly WP4-wide calls and several WP5 and cross- WP calls, as well as physical and remote Project Steering Board meetings, so as to unanimously make a final selection to be used in each respective Tier of the project. The selection also takes into account functional domains from WP2 as well as Patient Advisory Board suggestions and probe test outcomes.

1 D4.1 Device Selection

DoW Extract

Task 4.1: Device Evaluation, Selection and Refinement to Digitally Measure Functional Tasks (months 1 - 18, CERTH, OXF, JANSSEN, Takeda, Lilly, Novartis) The task entails two main subtasks, closely connected to one another, in order to select, bench test, extract features and realistically configure pilot devices according to instructions from WP2 - for WP5.

Device selection Device selection entails: a) the creation and rigorous update of a prioritised list of devices from the market and literature, b) initial selection from the list by ICT and clinical experts to meet functional task requirements of WP2, c) an initial laboratory assessment bench-test by experts and testers to adjust priorities and shortlist devices for the Tier 3 pilot, d) systematic refinement according to the Tier 3 pilot and delivery of the final selection to WP5. In detail, a) a list of devices with vendor, programmability (API), measurement type and scale, battery life, performance and comfort parameters will be organized, continuously populated and extended with both wearable and ambient devices. R & D devices, usually larger and less comfortable but also retail devices will be considered, such as presence, motion, object & utility usage, beacons, as they can be repurposed in a medical context for behavioural, cognitive and functional monitoring. Sources utilized will include the IMI ROADMAP review, interest groups such as those organized by C-Path, and databases like Vandrico. Given the list, b) experts will choose an initial set of devices by rating them based on knowledge and experience (including RADAR-CNS) with criteria such as: comfort, usefulness, functionality, unobtrusiveness, security and privacy features, programmability, robustness and accuracy to determine suitability. They will additionally assign a study tier (1, 2, or 3) to each device. The initial list will be c) bench-tested to pragmatically evaluate the above criteria in short trials with experts and testers, which will also provide standardized usability and user acceptance feedback1 that will be used to shortlist devices for the pilot in this Task (see next section). After this pilot, this task will iteratively d) incorporate full technological (precision, response time etc.) and clinical (suitability for the functional tasks, acceptance, usability2) feedback and metadata into the list and produce the final selection for WP5, ensuring that not only data science, but also a deep clinical understanding of the disease have contributed towards the outcome.

Lightweight Piloting of Candidate Devices with Health Age-matched An important part of device selection will be the evaluation of their suitability through a lightweight piloting exercise. This can use healthy age-matched controls and would take place in free-living environments. This short evaluation exercise would provide the opportunity to understand the user experience provided by each device and gather data from a real free-living situation. Ethical considerations will be explored, and guidance will be sought for the pilot. This application of this guidance will then be extended to cover the Tier 3 study in WP5 also to be undertaken by CERTH.

Related Deliverables:  D4.1.1 Device Selection Report for Tiers 1, 2 & 3 of the WP5 study (M12): Presents the device selection process, criteria and outcomes with the final choice of devices for pilots of all tiers.  D4.1.2 Device Selection Trials, clinical and ethical protocol for Tier 3 study to be used in WP5. (M18)

According to the DoW, this report, namely D4.1.1., refers to Device Selection activities while lightweight piloting will be reported in D4.1.2.

1 Davis, F. D., Bagozzi, R. P., & Warshaw, P. R. (1989). User acceptance of computer technology: a comparison of two theoretical models. Management science, 35(8), 982- 1003. 2 Brooke, J. (1996). "SUS: a "quick and dirty" usability scale". In P. W. Jordan, B. Thomas, B. A. Weerdmeester, & A. L. McClelland. Usability Evaluation in Industry, London: Taylor and Francis

2 D4.1 Device Selection

Contents

1 Introduction ...... 3 2 Search and Selection Methods...... 5 3 Results...... 5 3.1 Literature Search ...... 5 3.2 Tier 1 – Wearable Sensors ...... 7 Tier 1 – Final Selection ...... 8 3.3 Tier 2 – Home Sensors ...... 11 Tier 2 – Final Selection ...... 12 3.4 Tier 3 – Smart Home ...... 13 Tier 3 – Final Selection ...... 14 4 Conclusions ...... 16 5 References ...... 17 Appendix ...... 22 1. A – Tier 1: Wearable Market Search ...... 22 1. B – Tier 1: Wearables Listed Market Selection ...... 24 1. C – Tier 1: Apps Market Search ...... 25 1. D – Tier 1: Apps Listed Market Selection ...... 27 1. E – Tier 1: Wearable Cameras Market Search ...... 28 1. F – Tier 1: Wearable Sensors Categorization ...... 30 1. G – Tier 1: Wearable Literature Search...... 31 1. H – Tier 1: Apps Literature Search ...... 33 1. I – Tier 1: Wearable Cameras Literature Search...... 34 2. A – Tier 2 & 3: Smart Home Devices Market Search ...... 35 2. B – Tier 2 & 3: Smart Home Apps Market Search...... 39 2. C – Tier 2 & 3: Ambient Sensors Categories ...... 40 2. D – Tier 2 & 3: Literature Search Regarding IoT Wearable Sensors and Devices for Eldercare...... 41 2. E – Tier 2 & 3: Apps for Dementias Literature Search ...... 42 2. F – Tier 2 & 3: Review of Case Studies of IoT Wearable Sensors and Devices for Eldercare ...... 44 2. G – Tier 2 & 3: Smart Home Apps Literature Search ...... 45

1 Introduction

When assessing the level of difficulty and lack of function in daily living, we usually rely on feedback from family caregivers. This assessment may be influenced by subjective, imperfect recall and, thus, is not reliable to estimate the level of impairment across individuals. This absence of objective data could be mitigated following the advances in digital technology. Smartphones, currently owned by 9 out of 10 people, can help assess social behavior, via monitoring calls, SMS, or Internet browsing. Additionally, specialized smartphone applications may help assess cognition, step count or typing behavior. Accompanied by wearables, wristwatch or wristband sensors one can also measure stress level, heart rate, gait and more. Besides wearable, ambient sensors that disappear unobtrusively into one’s living environment can provide additional activity and behavioral monitoring. Pressure sensors fixed under the mattress can monitor sleep quality, duration and interruptions. GPS in the car can track driving navigation. Electricity sensors can monitor appliance usage to estimate cooking, doing chores, watching TV. The number of different devices and possibilities increases every year, resulting in a variety of factors that might affect developers, such as data heterogeneity, manufacturer standards and programming interfaces, but also end-users themselves, such as shapes, materials, battery life, design, functionality, precision, range and so on. All these parameters should be considered, when selecting proper devices for monitoring users, especially for the sensitive case of treating Alzheimer’s disease. Feedback from patients as well as their caregivers introduces a much- welcome end-user perspective in the selection process, introducing parameters that might be overlooked by researchers. Technology experts tend to select devices based only on more technical specs, so as to record the most appropriate signals with the highest granularity and precision. Still, the same devices might be widely uncomfortable, heavy and too complicated for the patients, bounding the project to failure. Thus, this selection should not be done without the participation of patients themselves. Patients and caregivers could stress attention to particular

3 D4.1 Device Selection

features tailored to the disease, memory and other functional impairments, as, for example, the necessity for a waterproof device or a sound or blinking light notification to remind users to charge the device. Then, an optimization process identifies the best compromise between the most technologically advanced and user-accepted devices to satisfy both parties. For this deliverable, we created a list of devices from the market and the literature. Every partner compiled their research on wearables, apps and devices (e.g. for smart-home usage) and presents their findings in this document. The literature and market reviews are considered and then a prioritized list is made, so the partners can finally reach an agreement of which devices / apps will be used in the forthcoming studies. The familiarity and previous usage or even trials by the partners with several of those devices helped adjust priorities, shortlist and finally select devices.

4 D4.1 Device Selection

2 Search and Selection Methods

In the following section, each partner suggested specific devices and apps available both in the Market as well as in the literature for the three Tiers, which will take place in WP5. An illustration of the pipeline of the present searching method is shown in Figure 1. First of all, we performed a targeted, “focused” literature search deepening in the field of wearable sensors and devices for elder care. The search included a wide range of applications, from chronic ailments, primarily Alzheimer’s disease (AD), other forms of dementia, Parkinson’s disease (PD), frailty and cardiovascular disease (CVD) to general eldercare and ambient assisted living (AAL). Internet of Things (IoT) technologies were categorized in wearables, as well as smart home sensors, cameras, microphones, and indoor and outdoor tracking. The literature search was published as Open Access and is available at [1]. From then on, we focused on commercial devices, either wearables or fixed, smart home devices, and apps, that are available to procure and use in the project trials. More specifically, we have searched for wearable devices and portable sensors for Tier 1 study which are available in the market or have been tested in research studies for similar approaches. Additionally, wearable cameras and a short study are presented, which took place in a previous research project. Moving forward, relevant fixed sensors are included, which can measure different modalities (i.e., cooking activity, TV use, driving activity etc.) and can be deployed at patients’ homes (Tier 2) since they are available in the market. In particular, we have included several studies which have deployed smart home devices to measure daily activity of elderly people. Finally, we present holistic approaches and smart home based technologies, which could be used as part of the Tier 3 study. Additionally, we have classified the sensors and technologies according to their ability to provide raw data, to their data types and based on the previous partners’ experience using a specific device for another study but also to the specific body functions they can measure.

Figure 1 – Pipeline of the Searching and Evaluation for the three Tiers

3 Results 3.1 Literature Search The literature search include a wide range of targeted research studies for IoT technology used in eldercare, was published as Open Access and is available online [2]. First we examined previous review studies and categorized them according their health focus (ailment), IoT technology used (wearables, smart home devices, cameras etc.) and review criteria they examined. Then we surveyed existing case studies and trials categorizing them according to health focus, IoT technology, aims (from assessment to fall detection and indoor positioning to intervention) and experimental evaluation. Particularly for experimental evaluation and trial results we further examined its duration, participants and outcome measures. Statistics drawn from this categorization allowed us to outline the current state-of-the-art, as well as trends and effective practices for the future of effective, accessible, and acceptable eldercare with technology. The study outline is shown on Figure 2.

5 D4.1 Device Selection

Figure 2 – Graphical Abstract of the Literature Review

Overal, we noticed that Alzheimer’s disease (AD) and dementia is the most prominent health focus of IoT eldercare applications followed by CVD, general care and frailty/fall detection, both in review studies (Figure 3) and individual case studies (Figure 4)

Figure 3 – Review studies according to their health focus. CVD, cardiovascular disease

Figure 4 – Case study papers according to their health focus.

The types of technology used include wearables in their majority and biometric sensors followed by ambient sensors (indoor positioning, environmental, smart home etc.), as shown on Figure 5.

Figure 5 – Review studies according to IoT technology devices are presented

6 D4.1 Device Selection

Another useful outcomes were the criteria examined in previous review studies, which we will also use to examined the candidate devices and apps, as shown on Figure 6. Further outcomes outside the scope of the device selection such as cohort size and characteristic are available in the paper.

Figure 6 – Categories of criteria examined in review studies

We also noticed (Figure 7) that the most prominent measure of interest is the technology’s accuracy followed by more human factors such as acceptance and user satisfaction, feedback etc.

Figure 7 – Case studies with evaluation according to their outcome measures

3.2 Tier 1 – Wearable Sensors We limited our research in the most well-known and compatible with the project’s needs devices, also according to all partners’ suggestions from the market (Table 4). Respectively, Table 10 refers to the same categories but as found in the literature. Devices measure step number, heart rate aggregates, users’ sleep patterns etc. They are presented along with parameters of significance for the study, such as the possibility to access raw data, their sensor types and their general concept of use. Access to raw data is of huge importance, as we need to visualize patients’ data in order to outline repetitive behaviours that will indicate possible AD. We also need to take into account their connection to functional domains relevant to early AD progression, as presented in WP2 of the RADAR-AD project. Table 5 presents a filtered list from all the device selection for Tier 1. The partners have searched the market for available devices, compatible with the Tiers. Then we consolidated every device and outlined the possible solutions that will work best for the needed studies. Therefore, a list is presented from which the final device selection will be decided. The Axivity AX3 can measure attributes of sleep, physical activity, and circadian rhythms, showing robust associations with 55 chronic disease outcomes in 100,000 UK Biobank participants. The device is capable of collecting raw 25Hz tri-axial acceleration data for at least 30 days without any charging. The proposed wear-time is all the time that participants are enrolled into the study. The required participant interaction is to replace the device once per month. The study team will sent a new fully charged device and provide a pre-paid envelope so that the old device can be sent back to the study- coordinating center. The partners also provided their market choices for applications as presented in Table 6. Patients can handle finances with Banking App or respond to reminders from Dementia Clock, allowing them to also prepare recipes in the kitchen. Furthermore, they can check their ability in task completion and combine it with smart home devices. Lastly, they can fill in questionnaires to achieve standardized scores. Table 7 presents a filtered list from all the app suggestions. The partners selected apps available from the market and then we collectively outlined their possible use and features, finally deciding which will work best for the project. Therefore, a list is presented from which the final app selection will be decided. The Mezurio app is suitable to this project as it protects its data strongly and it provides short memory tasks along with optional language tests. Altoida app requires an iPad, is both active and passive and deals with spatial memory and

7 D4.1 Device Selection

executive functions. Banking app simulates an ATM environment and helps with handling finances. Studies have shown that many cognitive impairments may require event segmentation. This can only be achieved with a wearable camera on a patient. By those means, the subject can practically be the scientists’ eyesight and they can study where the subject focuses, if he stares randomly, etc. This can be a good indicator of his/her current state. Those features are either sensory (such as color, sound, or movement) or conceptual (such as cause and effect interactions, and goals). In some forms of disease (e.g. schizophrenia) event segmentation is obstructed, something that defines its importance. If the patient’s eyesight also becomes the scientists’ eyesight, it will be easier to see the subject’s behavior through the day, where he randomly stares at, when he does not notice something etc. and diagnosis will be facilitated. In Table 8, we demonstrate detailed descriptions of wearable action cameras found in the market. Even though their purpose is mostly to record athletic activities, they enable us to access photos and videos from the user’s perspective and measure reaction time, route followed and other parameters that can provide useful insight for patients with Alzheimer’s disease. For the literature research we followed the same outline as the market search, but focusing on literature studies. Partners tried to find in the literature research studies with wearable devices, which tested the efficacy and improvement in patients with cognitive impairments and chronic diseases like Alzheimer’s disease. (Table 10). Devices with features that can help these patients are presented in the next tables, partnered with the literature findings on them. All partners categorized the wearable sensors according to their specific sensor type, measurement type and data rate. This categorization is necessary, so as to collect valuable information on the wearable sensors (Table 9). Sensors differentiate from accelerometers to pulse oximeters and thermal sensors; measurements are derived from units such as acceleration, respiration and body temperature and the data rates differentiate from very high to very low. Currently, healthcare providers are coping with ever-growing healthcare challenges, including ageing population, chronic diseases, cost of hospitalization and the risk of medical errors. Wearable patient monitoring systems are a potential solution for addressing some of these challenges by enabling advanced sensors, wearable technology, and secure and effective communication platforms between clinicians and the patients [3],[4]. Here, partners present their findings in wearable devices for Alzheimer patients. Studies found include a variety of sensors and many specific types of detection. The wearables in the literature are presented along with their aim, their wearing position and their general health focus in Table 10. In Table 11 we present all apps in the literature and, lastly, in Table 12 we outline the wearable cameras found in related publications.

Tier 1 – Final Selection Here are the final selections for Tier 1 (Table 1). Initial partners’ picks are filtered and a single option in each section is picked. Axivity AX3 scientific purpose is to continuously monitor 3D accelerometer data, which can possibly be linked to activities of daily living (ADLs), posture, physical activity level information. The continuous raw measurements will constitute a large valuable dataset for on-going and future research beyond the boundaries of the RADAR-AD project. The device constitutes the most affordable, reliable and comfortable raw accelerometer data logger, but on the other hand, it does not offer additional lifestyle measurements such as Heart Rate (HR) or Photoplethysmography (PPG). Our recommendation is to use the device to continuously monitor and store 3D acceleration in our dataset, valuable for on- going and future research. Still, due to its affordability, the device allows a second wearable to complement its role by measuring HR/PPG. Some studies using the device are found in the literature [5]–[7]. Fitbit Charge 3’s scientific purpose is to continuously monitor Heart Rate measurements. Steps, Calories and Sleep are extracted through a “blackbox” process by the manufacturer. However, they might still be insightful. It constitutes an affordable, reliable and comfortable data-logger for continuous/resting heart rate and steps, calories, distance, sleep despite the blackbox extraction. Also, it has no additional fees to access the data and is used unanimously in longitudinal research trials at home, but does not offer raw 3D accelerometer values; only the extracted steps, Sleep etc. Our recommendation is to use the device to continuously monitor and store heart rate measurements plus the blackbox- extracted Steps, Calories, Distance, Sleep. Despite the lack of raw 3D accelerometer values, due to its affordability, the device allows a second wearable to capture that. Some studies using the device are also found in the literature [8]– [11]. In order to select the two wearables, and ensure that the patients approve, a tailored probe was performed during the Patient Advisory Board meeting in Luxembourg by CERTH, OXF and JANBE. The patient representatives were presented with various wearables and polled about wearable features and patients’ preferences. They were also polled as to whether patients would accept the use of two wearables at once, considering e.g. the stigma, for which they responded positively. The detailed report can be found in the WP3 deliverables. In order to explore the acceptability and effectiveness of the wearable, wristband or wristwatch trackers, we have carefully designed the Human Factors and Technology Requirements Questionnaire (HFTRQ), available online, and applied it to a study group of 45 end-users distributed in three groups of 15

8 D4.1 Device Selection

participants each: healthcare professionals (HCP), caregivers and people with Mild Cognitive impairment (MCI) related to AD. The outcomes underlined several requirements of both groups that were taken into account for the selection. Findings include that HCP, caregivers and MCI participants are willing to adopt eHealth solutions based on wearables in order to assist in daily care through holistic and objective monitoring as well as the difficulties, peculiarities and priorities they are aspiring to alleviate through such systems. The results of the study have been published in IEEE ICHMS [12]. The patients will use their own smartphones during the trial and continue their daily life. Smartphones will be worn during lab assessment trials, namely the six minute walking test (6MWT), and will process accelerometer values to measure Gait parameters such as stride length, velocity etc. re-using the know-how developed in RADAR-CNS studies for the same purpose (but omitting the Faros device used in those studies). Additionally, the algorithms for extracting Gait are not “blackbox”, but rather open for the scientists to explore and modify and, lastly, effectiveness of measuring Gait in a 2MWT (now 6MWT) was established. On the other hand, RADAR-CNS studies have shown that smartphones with multiple apps active may present gaps during continuous data logging that may still be tolerable. Our recommendation is to use the device as the sole device for Gait measurement during the 6MWT as per the RADAR-CNS experience and then retain the raw accelerometer dataset for on-going and future studies as a valuable RADAR-AD asset. Some studies using the device are found in the literature [9], [10], [13]. Physilog device (Gait Up) is to be worn during lab assessment trials, namely the six one minute walking test (6MWT), normal and dual, and the timed- up and go test. The in-clinic gait assessment will be measured using three digital Physilog 6 devices combined with a validated gait-analysis software package3. Each device contains a triaxial accelerometer and gyroscope, barometer, and temperature sensor. The three devices will be attached securely to the dorsum of each foot and the lumbar spine. During the gait assessment tasks, the devices will stream the raw data in real-time, via Bluetooth, to the analysis software running on a Windows machine. The analysis software will calculate 26 meaningful metrics of gait during each task (walking speed, gait asymmetry, step variability, initial foot strike angle, etc.). These metrics have been fully validated against laboratory- based gold standards on various subject cohorts, including older adults and patients with Parkinson’s disease, cerebral palsy, and stroke. Once the assessment is complete, the software will provide full access to the raw data from the sensors as well as the processed gait metrics. Physilog has been used in multiple clinical studies and more than 400 publications4 processing accelerometer values to measure gait parameters such as stride length, velocity etc. re-using the know-how developed in RADAR-CNS studies for the same purpose (but omitting the Faros device used in those studies). It constitutes an affordable, reliable and comfortable raw logging/streaming device, it does not have any additional fees to extract gait parameters needed. Additionally, the algorithms for extracting gait will not be “blackbox”, but rather open for the scientists to explore and modify and, also, effectiveness of measuring GAIT in a 2MWT (now 16MWT) has been established in RADAR-CNS. On the other hand, RADAR-CNS studies have shown that smartphones with multiple apps active may present gaps during continuous data logging that may still be tolerable. Our recommendation is to use the device as the sole device for gait measurement during the 6MWT as per the RADAR-CNS experience and retain the raw accelerometer dataset for on-going and future studies as a valuable RADAR-AD asset. The Mezurio app, chosen by OXF partner, lets you complete interactive, scientifically valuable measurement tasks. It is designed to be used as part of a research study. It is a valuable part of that research as it allows the logging of data quickly, frequently, and from the comfort of one’s home. It is open, precise and secure and the data produced is carefully curated, strongly protected and easy to read. A study presenting this app is [14]. RADAR-base is a passive app collecting the following data streams from smartphone sensors: relative location, acceleration, gyration, magnetic field, step count, light, phone interaction status, application usage, bluetooth devices, and battery level. Altoida is a mobile phone and tablet-based digital biomarker platform that detects early and subtle micro-errors (accuracy) and micro-movements (latency), which has been shown to be useful in detecting MCI that will progress to dementia years in advance. A user-friendly exercise simulates a complex activity of daily living. Then the Altoida’s Neuro Motor Index (NMI), a performance score combining various data streams: voice data, hands micro-movements and micro-errors, gait micro-errors, posture changes, eye tracking, visuospatial navigation micro-errors etc., provides individually tailored prognostic information with the greatest ecological validity and scalability. Altoida’s NMI Medical Device has received DA class II medical device qualification from the FDA for the evaluation of perceptual, memory and functional impairment relevant to Activities of Daily living (ADL) and for assisting the diagnosis of MCI & AD in subjects between 55 and 95 years of age. NMI showed a diagnostic accuracy of 94% in predicting cognitive worsening of amyloid positive individuals who converted from MCI to AD after 5 years.

3 https://gaitup.com/science 4 https://gaitup.com/science

9 D4.1 Device Selection

Altoida's NMI has been used in different studies world-wide, as, for example, at a longitudinal study5 of 685 individuals. The Banking App, developed by CERTH, is an app to assess parameters related to the ability of managing finances during a clinical trial in the context of Dem@Care (http://demcare.eu). The cross-border trial with 190 participants in Thessaloniki, Greece and 100 in Nice, France revealed high accuracy (more than 80%) to discriminate between healthy, MCI and AD cognitive states, while the Banking App itself found statistically significant differences between the three groups in multiple parameters e.g. time to complete, correct input (PIN, amount, account), number of attempts etc. The app is implemented as a web application with a responsive User Interface which means it may be used on any PC, tablet or smartphone device. It simulates realistic usage of a bank ATM, prompting for a PIN code, amount and account number to transfer money to, in attempt to pay the bills. Meanwhile, it measures parameters related to the disease such as number of attempts, correctness of input, duration etc. The Banking App is recommended for usage in a lab setting under guidance of professionals as an assessment (not a monitoring tool), which is its purpose in RADAR-AD.

Device Price Modalities Data Comfort Partner Axivity AX3 o Manually o Battery life: Raw 3D extracted 28 days 120 £ OXF Accelerometer o Raw o Waterproof format o Light

o Steps o Calories Buffered in o Battery life: Fitbit o Distance the device, 28 days Charge 3 o Sleep streamed to o Waterproof segmentati smartphone, o Light on then o Touch 150 € KCL o Resting uploaded to screen and Fitbit cloud to display of Continuou be retrieved Clock, s heart via the open Steps, HR rate API. etc. o Exercise o Develope r friendly o Wireless SDKs & o 3D data APIs accelerom transfer Physilog o MATLAB eter o IP64 water (GaitUp) functions o 3D and dust 500 € and APIs, VUMC gyroscope resistant all o Barometric o MicroUSB available pressure for rapid for free sensors file o Gait and transfer. running analysis. o Accelerom o 12h Vicon eter continuous Autogra o Color Attempts recording pher sensor informed o Light o Magnetom decisions o Supplied 280 € eter about the software OXF o PIR sensor best time to for image o Temperatu capture a compilatio re picture. n into o GPS video Figure 8 – sensor Simulated ATM Transactions: o Response Banking Handling - time Active CERTH App finances o (in) correct amounts o Passwords etc. Short, daily episodic memory tasks in addition to Mezurio - optional Active Memory tasks OXF executive function and language tests.

5 https://clinicaltrials.gov/ct2/show/NCT02843529

10 D4.1 Device Selection

o iPad with pre- configured App and Web Active / NOVAR Altoida6 - Dashboard Memory tasks Passive TIS Login o Spatial memory, executive functions Table 1 – Final selection of devices and apps for Tier 1

3.3 Tier 2 – Home Sensors Elderly care at home is a matter of great concern, especially if the elderly live alone, since unforeseen circumstances might occur that affect their well-being. Low-cost, effective and reliable technology is essential for enhancing elderly care in independent living. Elderly care applications often demand real-time observation of the environment and the resident’s activities using an event-driven system. As an emerging area of research and development, it is necessary to explore the approaches of the elderly care system in the literature to identify current practices for future research directions [15]. There are already a number of technologies in use, including digital devices, smart sensors, and intelligent applications, which assist elderly people with their everyday needs in their own homes. Smart devices and sensors are installed in the house so they can monitor the subjects’ activity. Developing a strategy for an integrated technological solution would resolve many issues faced by elderly patients and would lead to improving their quality of life, health, and safety [16]. In this section, the partners search and present the most relevant and functional market products. We try to find the best solutions available, combining cost efficiency and functionality. Table 13 presents all suggestions for smart home devices in the market and Table 14 the suggestions for apps in the market. CERTH has implemented smart home devices in previous studies and will do so in Tier 2 of the RADAR-AD program. Plugwise Circle is a plug that streams data, manages the electric appliance (on/off) and can function with a number of daily living activities, such as cooking, watching TV etc. At an affordable cost it can be used for observation of different activities in daily living (ADLs). The Plugwise Scan detects motion and a general presence in a room, is affordable and reliable. Plugwise Sense is an environmental sensor that measures special events and is affordable, reliable and fast. Another device is CAO Gadgets Tag Sensor, which is the only sensor in the market to detect object movement, is affordable and effective but complicated to setup. Various door sensors can detect an open door or window, are very affordable and can be used to detect presence at home. Beacons (e.g. Estimote) are used for object or person proximity and for many ADLs. Sleep sensors placed under the mattress, such as Beddit or Nokia Sleep Sensor, are also being used. They are affordable, very reliable and unobtrusive and able to detect sleep stages, interruptions or time asleep per stage. Finally, Muse is a portable electroencephalographic (EEG) providing EEG metrics (after processing) and cognitive behavioral therapy (CBT) scores. Its affordability is a concern, due to its high cost. Home automation apps are continuously advancing, using leading edge technology, to observe more parameters and help predict outcomes more effectively. Digital devices, sensors, and intelligent applications are tools that can support seniors and allow better communication with their caregivers. The aim of this market review is to provide an up-to-date summary and serve as a reference for available technological solutions used to improve health and safety for people with Alzheimer’s disease. We have searched in the literature [15], [17]–[21] and categorized the ambient sensors according to their specific type, measurements, data format, aim and installation. This categorization is very effective as to collect all the valuable information on ambient sensors, as seen in Table 15. Measurements differ from motion detection, object identification etc.; data types can be categorical or numerical data; the same sensors are met in different installation sites, rooms, positioning and purpose. Fixed sensors are generally installed in the environment and attached in a fixed position. Often, they do not require being in direct contact with the user and are mostly unobtrusive, which can be an asset for the user. Examples of these types of sensors are video cameras, depth vision cameras, high quality microphones, motion sensors, etc. Sometimes, several sensors are installed in a common place usually called a smart environment [3]. It can be a single room, a house or an entire building. In a smart environment, the information gathered by those sensors can be used to understand the context of that environment in order to provide assistance, recommendations and services to the inhabitants [22]. In Table 16, we present the most relevant up-to-date publications highlighting the usage of fixed sensors found in the literature. They test specific modalities, such as passive infrared (PIR), bed, door, flood sensors or generally environmental,

6 Project Partner ALTOIDA: https://altoida.com/

11 D4.1 Device Selection

ambient or motor sensors. Their aim is also presented in this table, along with their outcomes. Each sensor measures a specific parameter and has a specific use. Alzheimer’s patients and their respective carers face difficulties and are in need of technological tools that can improve both the patient’s and the carer’s daily life. That level of observation and monitoring can also help to solve problems that the patients’ families and/or the caregivers face such as: the constant fear of losing their beloved ones at any moment they go out of the house without anyone knowing, and not remembering the important dates such as medical appointments, meal time etc. Accordingly, future apps projects will use modern technology to help caregivers take care of Alzheimer’s patients and maintain the level of their functions in the most effective possible way. In Table 17 we present the most relevant applications for the aid of elders and their caregivers, found in the literature. They test specific aspects of patients such as concentration, conversation, clinical use etc.

Tier 2 – Final Selection After a lot of research and many conversations between the partners, we outlined the budget and subsequently the devices for Tier 2 (Table 2). This study will take place in Amsterdam, Stockholm and London, with 40 participants. The available budget is €60.000 for 40 participants, which means €1500 for every participant, plus any remaining funds from Tier 1. The final selection resulted in Two Scenarios for this Tier. Both include the DREEM device and the OBD2 car GPS, but smart home selections are different in each: Specifically, the first scenario has more complex devices, meaning an RFID reader (€3.000) and 10 tags of other accelerometer-based sensors (<€100). This offers the ability to identify person and objects moved in the house, but also requires calibration processes in every new home, much more effort and the installations will be shared among the participants, meaning that we will install the sensors and complete all appropriate actions in a house, then remove everything and install again in the next house and so on. Finally, outcomes for algorithms to analyze and extraction of activities from these complex sensors are still in a very early stage. Little research has taken place using RFIDs for activity recognition, something that will result to unidentified risks in the whole procedure.

Figure 9. A graphic depiction of Tier 2 scenario 1 with complex devices The second scenario also includes the DREEM device and the OBD2 car GPS, but also simple devices: 2 PIR sensors, 2 plugs and 2 Door sensors (main entrance & fridge). For this scenario we have more state-of-the-art deployments to consolidate, but we still need the support of a Gateway (e.g. Raspberry Pi, Arduino) and development efforts, something that CERTH will undertake. Finally, by using this equipment we are not able to identify the patient vs. spouse/peer but we still have an important insight of what happens in the observed environment.

Figure 10. A graphic depiction of Tier 2 Scenario 2 with simple devices After presenting the two scenarios, emphasizing in pros and cons at the WP4 Device Selection track meeting, the WP4-wide meeting, the Cross-WP meeting and the Project Steering Board meeting, we unanimously made the decision to proceed with Scenario two. It was also decided to reserve complex devices and bigger quantities of the simple devices, coupled with innovative algorithms to process the additional and more diverse data, for Tier 3. The indoor localization for an object or a person can be obtained with a RFID sensor, such as Atlas RFID Solutions, however this was considered candidate for selection with at the high cost of $1,585. Another RFID reader/transponder (along with its tags) was suggested as a cheaper alternative for the testing scale of our trial, manufactured by SparkFun Company. Finally, an accelerometer is considered as a possible solution for activity recognition such as BluEpyc BE-DSK-A-BLE.1 Bluetooth BLE Disk Beacon, at the cost of $2,592 (100 beacons) or MetaWear BLE Sensors with the same cost approximately, depending on the kit selected. The DREEM device costs €500 and has the advantage of re-usage among patients and different Tiers, which means it is would also be available for the next Tier. Another device proposed is the car GPS, which allows us to have a visual in the subject’s driving, the way he maneuvers the car and their general driving ability. Finally, inside each subject’s house, smart plugs, door sensors and PIR sensors will be placed. More specifically, a minimum of two sensors will be installed (fridge door, entrance door) and 2-3 presence sensors in places like the bathroom, living room or kitchen. If sensors that can be also used in the next Tier are chosen, it

12 D4.1 Device Selection

means there would be more funds available for the next one, so we could also use more indoor localizations sensors and also consider RFID sensors. Data Sensor Smart Access Device data types WP2 Domain Part Home (Strea Characteristic (Modalities – Concept ner Device m or s ) Log) DREEM 599 € Portable EEG metrics (unlimited JAN Log EEG for and Sleep access to raw BE sleep Quality data)

OBD2 car Bluetooth diagnostic GPS JAN Stream Bluetooth interface Driving ability BE scanner (Android) On-off, ADLs e.g. Plugwise o Affordable, works any Cooking, Circle reliable and power house chores, (Plug) fast (40€ supplied watching TV – CER Stream per appliance, combined with TH appliance) even non- presence Useful for socket sensors and ADLs appliances processing o Affordable, Plugwise reliable and Scan fast (Motion) Presence in ADLs e.g. o Integrates CER Stream a room (IR bathroom with plugs TH motion) visits easily o Useful for ADLs At-home or o Very taking a walk, Various Affordable doing outside Open door / CER Door - o Useful for chores (after window TH Sensors presence at processing, home combined with other sensors) Table 2 – Final selection of sensors and devices for Tier 2

3.4 Tier 3 – Smart Home The smart home could contain different sensors (movement sensors, door entry point, taps, kettles, cookers sensors, etc.) to determine different classes of context, which would help to identify patterns of use and movement, and eventually allow the categorization of the user’s behavior. When the behavioral pattern is learned, any anomalous behavior could then be detected. The most important factor in designing a smart environment for the elderly is that the technology should not interfere with the normal activities of the patient. Thus, all devices should operate autonomously. We intend to use only low cost and readily available sensors which could be installed by the users themselves or their informal carers. In this section, the partners present their findings in smart home holistic approaches. Multiple modalities are examined (e.g. temperature control, appliances, wireless remote monitoring solutions) and all have to function as a remotely controlled unit. Home automation and smart homes are the two ambiguous terms used in reference to a wide range of solutions for monitoring, controlling, and automating functions in a home. The smart home system requires a smartphone application or web portal as a user interface, to interact with an automated system. The scope of this study includes an analysis of the devices that can be controlled by switches, timers, sensors, and remote controllers, apart from other control devices. Globally, the increasing importance toward the need to counter security issues is anticipated to fuel up the demand growth for smart and connected homes over the forecast period. Moreover, the introduction of innovative wireless technologies, including Heating, Ventilation and Air Conditioning (HVAC) Controller, security and access regulators and entertainment controls, is expected to foster market growth. Furthermore, the recent advancements in IoT resulted in price drops of sensors and processors and are expected to encourage manufacturers to promote automation in the household sector [23]. In this Tier 10 persons from Tier 1 will be evaluated in a smart home and the budget is €40.000. Therefore, we have screened for relevant smart home devices in the market for Tier 2 and in Table 3 we present the final suggestions for Smart Home including information concerning data access, sensor types and their general concept. A problem that occurred is that sensors such as PIR, plugs and door sensors do not identify the exact person using an object. To that end, we suggest using a Raspberry Pi with an Arduino, programmed by CERTH. For the indoor localization we proposed the use of RFID sensors or Bluetooth and GPS sensors. All of our suggestions can be found in Table 13, together with Tier 2 devices.

Technology has been exponentially growing for the smart home sector and smart home technologies now consist of devices and services that are connected to the internet and to each other. The user can access and manage them remotely using mobile devices. Subsequently, one can monitor and control their home

13 D4.1 Device Selection

devices remotely, offering facilitation and advanced security. The adoption of the IoT has already captured smart home enthusiasts. New smart home technologies that are appearing and leading companies are embracing leading edge technology. In the past five years, voice-based technologies have experienced rapid growth in the consumer market. Market leaders such as Amazon, Google, Apple, and Samsung continue to enhance their product line. In the future, we expect to have a number of voice-activated smart devices in our homes. We might no longer use light switches but instead giving voice commands asking to turn and off light. The same will be with HVAC systems, TV, and any other device [24]. Nowadays, there are several air monitoring products on the market that send messages about the air in your home. These devices can be set up at smart home and improve the quality of the air in the interiors of the home. One of the biggest concerns for most homeowners is water leaks and its consequences. Especially areas with frequent floods and hurricanes suffer from the damage caused by a leakage event. Unfortunately, at the time we notice the leak it is already too late and the house could already be damaged. For this reason, leak detector systems and valve controllers sounds promising. Home security solutions allow residents to monitor each room. With sensors and alarms being installed throughout the house, it is possible to discover any changes in the building. Whether it is a smoke detection, open doors or motion detection, the smart home system notices the user about those changes. Unlike most home security systems, all-in-one-systems require minimal installation. Each device comprises of live- streaming cameras, motion sensors, arming and disarming functionality, and integrated sirens [24]. Here, we demonstrate apps found in the market. Their functionality is to keep all smart home devices in a specific environment so the user can manage them all in once, from the same app. To that end, we have screened for relevant apps connected to smart home devices in the market for Tier 2, and then we selected the final ones, which will also be deployed in Tier 3. Therefore, in Table 3 (together with Tier 2 apps), we present the final suggestions for Smart Home apps including information concerning the modalities, sensor types and their general concept. The technology of Smart Homes, as an instance of ambient assisted living technologies, is designed to assist the homes’ residents accomplishing their daily- living activities and thus having a better quality of life while preserving their privacy [3]. A Smart Home system is usually equipped with a collection of inter-related software and hardware components to monitor the living space by capturing the behavior of the resident and understanding his activities. By doing so the system can inform about risky situations and take actions on behalf of the resident to his satisfaction [19]. The literature has a lot of survey papers about wearable technology and taxonomy, but also many studies for Smart Home sensors and devices, shown in Table 18. Many examples are included, such as in-house IoT sensors, smart home and health care projects, robotic service platforms and human machine interfaces etc. The studies are presented along with the sensors included, their health focus and finally their aim.

In the ideal version of a wired future, all devices in smart homes communicate with one another seamlessly. Smart home technology based on IoT has changed human life by providing connectivity to everyone regardless of time and place. Home automation systems have become increasingly sophisticated in recent years. These systems provide infrastructure and methods to exchange all types of appliance information and services. Smart homes are automated buildings with installed detection and control devices, such as air conditioning and heating, ventilation, lighting, hardware, and security systems. These modern systems, which include switches and sensors that communicate with a central axis, are sometimes called “gateways”. These “gateways” are control systems with a user interface that interacts with a tablet, mobile phone, or computer; the network connectivity of these systems is managed by IoT. Since 2010, researchers have analyzed IoT-based smart home applications using several approaches. Regardless of their category, existing research articles focus on the challenges that hinder the full utilization of smart home IoT applications and provide recommendations to mitigate these problems. Research on smart home applications is dynamic and diverse [25]. In Table 19, the partners present their suggestions for smart home managing apps, found in the literature. Tier 3 – Final Selection In Tier 3, the final selection is also following a lot of research and discussions by the partners. For Tier 3 the budget available is €40.000 and it will take place in the smart home of CERTH. This allows us to fill the smart home with various sensors and tags, so we can acquire a lot of important data and analyze many aspects of the subjects’ way of living. An amount of 10 participants will be staying in the smart home, but the sensors will be installed once and used in all of the subjects, so we can use the whole budget. The specific sensors that will be used are RFID and indoor localization. We will use again the Tier 2 devices but the quantity of simple devices will be greater. We will add PIR sensors in every room, door sensors in all of the smart home’s main doors and smart plugs in every appliance (Table 3). Additionally, we can install and calibrate (once) some complex accelerometers for objects or/and expensive RFID sensors to localize the person and the objects moved each time in the house. This corresponds to a combination and extension of the two scenarios for Tier 2 that will be complemented and explored here.

14 D4.1 Device Selection

Figure 11. A graphic depiction of the Tier 3 devices complex and simple. Data Smart access Partne Home Sensor types WP2 Connection (Stream / r App Log) DREEM Portable EEG for EEG metrics and Sleep JANB Log sleep Quality E

OBD2 car JANB GPS Stream Bluetooth Driving Ability E

Plugwi se Circle On-off, works any ADLs e.g. Cooking, power supplied house chores, watching (Plug) CERT Stream appliance, even TV – combined with H non-socket presence sensors and appliances processing

Plugwi se Scan (Motion Presence in a ADLs e.g. bathroom CERT ) Stream room (IR motion) visits H

At-home or taking a Variou walk, doing outside s Door Open door / CERT - chores (after Sensor window H processing, combined s with other sensors) Table 3 – Final selection of sensors and devices for Tier 3

15 D4.1 Device Selection

4 Conclusions

In this deliverable we present the complete procedure for the device selection for the WP4 section of the RADAR-AD project. The whole process is presented, from the initial literature search, to the market search, the filtration of the first choices and the final selection for each Tier. We also present some more technical features of the devices, the sensors types and general filters for selection. Initially, literature research examined a wide range of IoT technology, wearables, smart home devices and apps used in various eldercare scenarios, including AD. The research was published and is available as Open Access [1]. Then, we also constructed a tailored questionnaire to identify the human factors and technological requirements of elders, caregivers and medical professionals when using technology and performed an explorative study to drive the selection process [12]. As for the final selection for each Tier, all the partners contributed and after many conversations have been held the decision was made as to which devices will have the most benefits for the whole Tier. For Tier 1 FitBit Charge 3 and Axivity AX 3 were selected to be placed on the wrist of each subject. The difference between these 2 devices is that the second one provides raw data for gait as it implements a 3D Accelerometer. The first device also measures sleep, but we cannot obtain the raw data necessary for the research. Physilog device measures gait and will be used during a short gait test. The apps chosen are Mezurio, Altoida and Banking App developed by CERTH. The first two are dealing with memory tasks and the latter provides information about the subject’s ability of handling finances, simulating an ATM. For Tier 2, DREEM device was unanimously proposed as it has unlimited access to raw data, provides EEG metrics and the subject’s sleep quality. DREEM device costs €500 and has the advantage of re-usage among patients and different Tiers, which means it is also available for the next tier. The car GPS allows us to have a visual in the subject’s general driving ability. Inside each subject’s house, smart plugs, door and PIR sensors will be placed. Finally, for Tier 3 we selected the Tier 2 devices but functioning in the CERTH Smart Home. Each device will have a smart plug, each door will have an RFID sensor and presence sensors will this time be increased in quantity in each room. Last but not least, the appendix section provides the tables presenting all the detailed information about all the products examined and their characteristics and also all of the literature sources.

16 D4.1 Device Selection

5 References

[1] T. Stavropoulos, S. Andreadis, L. Mpaltadoros, S. Nikolopoulos, and I. Kompatsiaris, “Wearable Sensors and Smartphone Apps as Pedometers in eHealth: a Comparative Accuracy, Reliability and User Evaluation,” in 1st IEEE International Conference on Human-Machine Systems, 2020. [2] T. G. Stavropoulos, A. Papastergiou, L. Mpaltadoros, S. Nikolopoulos, and I. Kompatsiaris, “Iot wearable sensors and devices in elderly care: A literature review,” Sensors (Switzerland), vol. 20, no. 10, 2020. [3] T. Stavropoulos, S. Nikolopoulos, and I. Kompatsiaris, “Sensors in Everyday Objects for Dementia Care - HealthManagement.org,” 2019. [4] M. M. Baig, H. GholamHosseini, A. A. Moqeem, F. Mirza, and M. Lindén, “A Systematic Review of Wearable Patient Monitoring Systems – Current Challenges and Opportunities for Clinical Adoption,” J. Med. Syst., vol. 41, no. 7, p. 115, Jul. 2017. [5] A. Mehmood Khan, G. Kalkbrenner, and M. Lawo, “Recognizing Physical Training Exercises Using the Axivity Device.” [6] A. Doherty et al., “Large Scale Population Assessment of Physical Activity Using Wrist Worn Accelerometers: The UK Biobank Study.,” PLoS One, vol. 12, no. 2, p. e0169649, 2017. [7] R. McArdle, B. Galna, A. Thomas, and L. Rochester, “CONTINUOUS MONITORING OF GAIT: WHAT CAN IT TELL US ABOUT DEMENTIA?,” Alzheimer’s Dement., vol. 14, no. 7, pp. P754–P755, Jul. 2018. [8] K. M. Diaz et al., “Fitbit®: An accurate and reliable device for wireless physical activity tracking.,” Int. J. Cardiol., vol. 185, pp. 138–40, Apr. 2015. [9] S. S. Paul et al., “Validity of the Fitbit activity tracker for measuring steps in community-dwelling older adults,” BMJ Open Sport Exerc. Med., vol. 1, no. 1, Oct. 2015. [10] T. Banerjee, M. Peterson, Q. Oliver, A. Froehle, and L. Lawhorne, “Validating a Commercial Device for Continuous Activity Measurement in the Older Adult Population for Dementia Management.,” Smart Heal. (Amsterdam, Netherlands), vol. 5–6, pp. 51–62, Jan. 2018. [11] S. P. Wright, T. S. Hall Brown, S. R. Collier, and K. Sandberg, “How consumer physical activity monitors could transform human physiology research,” American Journal of Physiology - Regulatory Integrative and Comparative Physiology, vol. 312, no. 3. American Physiological Society, pp. R358–R367, Mar-2017. [12] T. G. Stavropoulos, I. Lazarou, D. Strantsalis, G. Koumanakos, M. Frouda, and M. Tsolaki, “Human Factors and Requirements of People with Mild Cognitive Impairment , their Caregivers and Healthcare Professionals for eHealth Systems with Wearable Trackers,” in 1st IEEE International Conference on Human-Machine Systems, 2020. [13] J. M. Perez-Macias, H. Jimison, I. Korhonen, and M. Pavel, “Comparative assessment of sleep quality estimates using home monitoring technology,” in 2014 36th Annual International Conference of the IEEE Engineering in Medicine and Biology Society, EMBC 2014, 2014, pp. 4979–4982. [14] C. Lancaster, I. Koychev, J. Blane, A. Chinner, L. Wolters, and C. Hinds, “The Mezurio smartphone application: Evaluating the feasibility of frequent digital cognitive assessment in the PREVENT dementia study,” medRxiv, p. 19005124, 2019. [15] M. Z. Uddin, W. Khaksar, and J. Torresen, “Ambient Sensors for Elderly Care and Independent Living: A Survey.,” Sensors (Basel)., vol. 18, no. 7, Jun. 2018. [16] P. Maresova et al., “Technological Solutions for Older People with Alzheimer’s Disease: Review,” Curr. Alzheimer Res., vol. 15, no. 10, pp. 975–983, Apr. 2018. [17] P. Rashidi and A. Mihailidis, “A Survey on Ambient-Assisted Living Tools for Older Adults,” IEEE J. Biomed. Heal. Informatics, vol. 17, no. 3, pp. 579–590, May 2013. [18] M. H. Ur Rehman, C. S. Liew, T. Y. Wah, J. Shuja, and B. Daghighi, “Mining personal data using smartphones and wearable devices: A survey,” Sensors (Switzerland), vol. 15, no. 2. MDPI AG, pp. 4430–4469, 2015. [19] M. Amiribesheli, A. Benmansour, and A. Bouchachia, “A review of smart homes in healthcare,” J. Ambient Intell. Humaniz. Comput., vol. 6, no. 4, pp. 495–517, Aug. 2015. [20] D. Ding, R. A. Cooper, P. F. Pasquina, and L. Fici-Pasquina, “Sensor technology for smart homes,” Maturitas, vol. 69, no. 2. pp. 131–136, Jun- 2011. [21] Y. Zang, F. Zhang, C. A. Di, and D. Zhu, “Advances of flexible pressure sensors toward artificial intelligence and health care applications,” Materials Horizons, vol. 2, no. 2. Royal Society of Chemistry, pp. 140–156, 2015. [22] E. Garcia-Ceja, M. Riegler, T. Nordgreen, P. Jakobsen, K. J. Oedegaard, and J. Tørresen, “Mental health monitoring with multimodal sensing and machine learning: A survey,” Pervasive and Mobile Computing. 2018.

17 D4.1 Device Selection

[23] “Smart Homes Market | Growth, Trends, and Forecast (2019 - 2024).” . [24] “Fresh Trends in Smart Home Apps’ Development | Agilie app development company blog.” . [25] M. Alaa, A. A. Zaidan, B. B. Zaidan, M. Talal, and M. L. M. Kiah, “A review of smart home applications based on Internet of Things,” Journal of Network and Computer Applications, vol. 97. Academic Press, pp. 48–65, Nov-2017. [26] K. K. B. Peetoom, M. A. S. Lexis, M. Joore, C. D. Dirksen, and L. P. De Witte, “Literature review on monitoring technologies and their outcomes in independently living elderly people,” Disability and Rehabilitation: Assistive Technology, vol. 10, no. 4. Informa Healthcare, pp. 271–294, Jul-2015. [27] P. Cedillo, C. Sanchez, K. Campos, and A. Bermeo, “A Systematic Literature Review on Devices and Systems for Ambient Assisted Living: Solutions and Trends from Different User Perspectives,” 2018 5th Int. Conf. eDemocracy eGovernment, ICEDEG 2018, pp. 59–66, 2018. [28] L. Piwek, D. A. Ellis, S. Andrews, and A. Joinson, “The Rise of Consumer Health Wearables: Promises and Barriers,” PLoS Med., vol. 13, no. 2, pp. 1–9, 2016. [29] M. Haghi, K. Thurow, I. Habil, R. Stoll, and M. Habil, “Wearable Devices in Medical Internet of Things,” Heal. Informatics Res., vol. 23, no. 1, pp. 4– 15, 2017. [30] S. Seneviratne et al., “A Survey of Wearable Devices and Challenges,” IEEE Commun. Surv. Tutorials, vol. 19, pp. 2573–2620, 2017. [31] D. Surendran, J. Janet, D. Prabha, and E. Anisha, “A Study on devices for assisting Alzheimer patients,” Proc. Int. Conf. I-SMAC (IoT Soc. Mobile, Anal. Cloud), I-SMAC 2018, pp. 620–625, 2019. [32] S. L. Chen, H. Y. Lee, C. A. Chen, H. Y. Huang, and C. H. Luo, “Wireless body sensor network with adaptive low-power design for biometrics and healthcare applications,” IEEE Syst. J., vol. 3, no. 4, pp. 398–409, 2009. [33] S. Patel, H. Park, P. Bonato, L. Chan, and M. Rodgers, “art%3A10.1186%2F1743-0003-9-21,” J. Neuroengineering Rehabil., pp. 1–17, 2012. [34] J. S. Talboom and M. J. Huentelman, “Big data collision: the internet of things, wearable devices and genomics in the study of neurological traits and disease.,” Hum. Mol. Genet., vol. 27, no. R1, pp. R35–R39, May 2018. [35] D. V. Dimitrov, “Medical internet of things and big data in healthcare,” Healthc. Inform. Res., vol. 22, no. 3, pp. 156–163, 2016. [36] N. Scarpato, A. Pieroni, L. Di Nunzio, and F. Fallucchi, “E-health-IoT universe: A review,” Int. J. Adv. Sci. Eng. Inf. Technol., vol. 7, no. 6, pp. 2328–2336, 2017. [37] S. Barri Khojasteh, J. R. Villar, E. de la Cal, V. M. González, J. Sedano, and H. R. Yazg̈ an, “Evaluation of a Wrist-Based Wearable Fall Detection Method,” 2018, pp. 377–386. [38] R. J. Ellis et al., “A Validated Smartphone-Based Assessment of Gait and Gait Variability in Parkinson’s Disease,” 2015. [39] A. Weiss et al., “The transition between turning and sitting in patients with Parkinson’s disease: A wearable device detects an unexpected sequence of events,” Gait Posture, vol. 67, pp. 224–229, Jan. 2019. [40] R. Mc Ardle et al., “Gait in Mild Alzheimer’s Disease: Feasibility of Multi- Center Measurement in the Clinic and Home with Body-Worn Sensors: A Pilot Study,” J. Alzheimers. Dis., vol. 63, no. 1, pp. 331–341, 2018. [41] L. Costa et al., “Application of Machine Learning in Postural Control Kinematics for the Diagnosis of Alzheimer’s Disease,” 2016. [42] H. Zhou, M. Sabbagh, R. Wyman, C. Liebsack, M. E. Kunik, and B. Najafi, “Instrumented Trail-Making Task (iTMT) to Differentiate Persons with No Cognitive Impairment, Amnestic Mild Cognitive Impairment, Alzheimer’s Disease-Proof of Concept Study.” [43] Y.-L. Hsu et al., “Gait and Balance Analysis for Patients With Alzheimer’s Disease Using an Inertial-Sensor-Based Wearable Instrument,” IEEE J. Biomed. Heal. Informatics, vol. 18, no. 6, pp. 1822–1830, Nov. 2014. [44] S. Abbate, M. Avvenuti, and J. Light, “Usability Study of a Wireless Monitoring System among Alzheimer’s Disease Elderly Population,” 2014. [45] A. M. Pot, B. M. Willemse, and S. Horjus, “A pilot study on the use of tracking technology: feasibility, acceptability, and benefits for people in early stages of dementia and their informal caregivers.,” Aging Ment. Health, vol. 16, no. 1, pp. 127–134, 2012. [46] W. L. Giggins O, Clay I, “Physical Activity Monitoring in Patients with Neurological Disorders: A Review of Novel Body-Worn Devices.,” Digit. Biomarkers, vol. 1, pp. 14–42, 2017. [47] D. Lie et al., “A 2.4GHz Non-Contact Biosensor System for Continuous Monitoring of Vital-Signs,” in Telemedicine Techniques and Applications, InTech, 2011. [48] G. Wu and S. Xue, “Portable preimpact fall detector with inertial sensors.,” IEEE Trans. Neural Syst. Rehabil. Eng., vol. 16, no. 2, pp. 178–83, Apr. 2008.

18 D4.1 Device Selection

[49] C. F. Lai, S. Y. Chang, H. C. Chao, and Y. M. Huang, “Detection of cognitive injured body region using multiple triaxial accelerometers for elderly falling,” IEEE Sens. J., vol. 11, no. 3, pp. 763–770, 2011. [50] C. Setz, B. Arnrich, J. Schumm, R. La Marca, G. Tröster, and U. Ehlert, “Discriminating stress from cognitive load using a wearable eda device,” IEEE Trans. Inf. Technol. Biomed., vol. 14, no. 2, pp. 410–417, Mar. 2010. [51] S. S. Aljehani, R. A. Alhazmi, S. S. Aloufi, B. D. Aljehani, and R. Abdulrahman, “iCare: Applying IoT Technology for Monitoring Alzheimer’s Patients,” in 2018 1st International Conference on Computer Applications & Information Security (ICCAIS), 2018, pp. 1–6. [52] E. Z. Pirani, F. Bulakiwala, M. Kagalwala, M. Kalolwala, and S. Raina, “Android Based Assistive Toolkit for Alzheimer,” in Procedia Computer Science, 2016, vol. 79, pp. 143–151. [53] M. Karakaya, G. Şengül, and A. Bostan, “REMOTELY MONITORING ACTIVITIES OF THE ELDERS USING SMART WATCHES EEG Source Localization View project REMOTELY MONITORING ACTIVITIES OF THE ELDERS USING SMART WATCHES,” 2017. [54] A. R. Silva, M. S. Pinho, L. Macedo, C. Moulin, S. Caldeira, and H. Firmino, “It is not only memory: effects of sensecam on improving well- being in patients with mild alzheimer disease,” Int. Psychogeriatrics, vol. 29, no. 5, 2017. [55] E. Woodberry, G. Browne, S. Hodges, P. Watson, N. Kapur, and K. Woodberry, “The use of a wearable camera improves autobiographical memory in patients with Alzheimer’s disease,” Memory, vol. 23, no. 3, pp. 340–349, 2015. [56] M. C. Allé, A. Giersch, J. Potheegadoo, N. Meyer, J.-M. Danion, and F. Berna, “From a Lived Event to Its Autobiographical Memory: An Ecological Study Using Wearable Camera in Schizophrenia,” Front. Psychiatry, vol. 10, Oct. 2019. [57] F. M. Li, D. L. Chen, M. Fan, and K. N. Truong, “FMT,” Proc. ACM Interactive, Mobile, Wearable Ubiquitous Technol., vol. 3, no. 3, pp. 1–25, Sep. 2019. [58] E. Berry et al., “The use of a wearable camera, SenseCam, as a pictorial diary to improve autobiographical memory in a patient with limbic encephalitis: A preliminary report,” Neuropsychol. Rehabil., vol. 17, no. 4– 5, pp. 582–601, Aug. 2007. [59] S. Hodges, E. Berry, and K. Wood, “SenseCam: a wearable camera that stimulates and rehabilitates autobiographical memory.,” Memory, vol. 19, no. 7, pp. 685–96, Oct. 2011. [60] R. Brindley, A. Bateman, and F. Gracey, “Exploration of use of SenseCam to support autobiographical memory retrieval within a cognitive- behavioural therapeutic intervention following acquired brain injury,” Memory, vol. 19, no. 7, pp. 745–757, 2011. [61] Z. Wang, Z. Yang, and T. Dong, “A review of wearable technologies for elderly care that can accurately track indoor position, recognize physical activities and monitor vital signs in real time,” Sensors (Switzerland), vol. 17, no. 2, 2017. [62] S. Blackman et al., “Ambient Assisted Living Technologies for Aging Well: A Scoping Review,” J. Intell. Syst, vol. 25, no. 1, pp. 55–69, 2016. [63] M. M. Baig, S. Afifi, H. GholamHosseini, and F. Mirza, “A Systematic Review of Wearable Sensors and IoT-Based Monitoring Applications for Older Adults – a Focus on Ageing Population and Independent Living,” J. Med. Syst., vol. 43, no. 8, 2019. [64] V. G. Spasova and I. Iliev, “A survey on automatic fall detection in the context of ambient assisted living systems,” 2014. [65] M. Ienca et al., “Intelligent Assistive Technology for Alzheimer’s Disease and Other Dementias: A Systematic Review,” J. Alzheimer’s Dis., vol. 56, no. 4, pp. 1301–1340, 2017. [66] M. Alwan et al., “Validation of rule-based inference of selected independent activities of daily living.,” Telemed. J. E. Health., vol. 11, no. 5, pp. 594–9, Oct. 2005. [67] A. Ariani, S. J. Redmond, D. Chang, and N. H. Lovell, “Simulated unobtrusive falls detection with multiple persons.,” IEEE Trans. Biomed. Eng., vol. 59, no. 11, pp. 3185–96, Nov. 2012. [68] A. Bamis, D. Lymberopoulos, T. Teixeira, and A. Savvides, Towards Precision Monitoring of Elders for Providing Assistive Services. 2008. [69] K.-Y. Chung, K. Song, K. Shin, J. Sohn, S. H. Cho, and J.-H. Chang, “Noncontact Sleep Study by Multi-Modal Sensor Fusion.,” Sensors (Basel)., vol. 17, no. 7, Jul. 2017. [70] T. Guettari et al., “Multimodal localization in the context of a medical telemonitoring system,” pp. 3835–3838, 2010. [71] J. M. Kinney, C. S. Kart, L. D. Murdoch, and C. J. Conley, “Striving to Provide Safety Assistance for Families of Elders:The SAFE House Project,” Dementia, vol. 3, no. 3, pp. 351–370, 2004. [72] A. Lotfi, C. Langensiepen, S. M. Mahmoud, and M. J. Akhlaghinia, “Smart homes for the elderly dementia sufferers: Identification and prediction of

19 D4.1 Device Selection

abnormal behaviour,” J. Ambient Intell. Humaniz. Comput., vol. 3, no. 3, pp. 205–218, Sep. 2012. [73] M. Rantz, M. Skubic, S. Miller, and J. Krampe, “Using technology to enhance aging in place,” in Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), 2008, vol. 5120 LNCS, pp. 169–176. [74] J. Van Hoof, H. S. M. Kort, P. G. S. Rutten, and M. S. H. Duijnstee, “Ageing-in-place with the use of ambient intelligence technology : Perspectives of older users,” Int. J. Med. Inform., vol. 80, no. 5, pp. 310– 331, 2011. [75] F. Zhou, J. Jiao, S. Chen, and D. Zhang, “A case-driven ambient intelligence system for elderly in-home assistance applications,” IEEE Trans. Syst. Man Cybern. Part C Appl. Rev., vol. 41, no. 2, pp. 179–189, Mar. 2011. [76] N. Zouba, F. Brémond, M. Thonnat, M. T. Multisensor, and F. Bremond, “Multisensor Fusion for Monitoring Elderly Activities at Home,” 2009. [77] L. Rachakonda, S. P. Mohanty, E. Kougianos, K. Karunakaran, and M. Ganapathiraju, “Smart-pillow: An IoT based device for stress detection considering sleeping habits,” in Proceedings - 2018 IEEE 4th International Symposium on Smart Electronic Systems, iSES 2018, 2019, pp. 161–166. [78] J. M. Goodman-Casanova et al., “TV-based assistive integrated service to support European adults living with mild dementia or mild cognitive impairment (TV-AssistDem): study protocol for a multicentre randomized controlled trial.,” BMC Geriatr., vol. 19, no. 1, p. 247, Sep. 2019. [79] D. Welsh et al., “Ticket to talk: Supporting conversation between young people and people with dementia through digital media,” in Conference on Human Factors in Computing Systems - Proceedings, 2018, vol. 2018- April. [80] D. Shibata, M. Miyabe, S. Wakamiya, A. Kinoshita, K. Ito, and E. Aramaki, “VocabChecker: Measuring language abilities for detecting early stage dementia,” in International Conference on Intelligent User Interfaces, Proceedings IUI, 2018. [81] M. Atee, K. Hoti, and J. D. Hughes, “A Technical Note on the PainChekTM System: A Web Portal and Mobile Medical Device for Assessing Pain in People With Dementia.,” Front. Aging Neurosci., vol. 10, p. 117, 2018. [82] C. Tyack, P. M. Camic, M. J. Heron, and S. Hulbert, “Viewing Art on a Tablet Computer: A Well-Being Intervention for People With Dementia and Their Caregivers.,” J. Appl. Gerontol., vol. 36, no. 7, pp. 864–894, 2017. [83] R. E. Docking, M. Lane, and P. A. Schofield, “Usability Testing of the iPhone App to Improve Pain Assessment for Older Adults with Cognitive Impairment (Prehospital Setting): A Qualitative Study.,” Pain Med., vol. 19, no. 6, pp. 1121–1131, 2018. [84] Y. P. Huang, A. Singh, S. Chen, F. J. Sun, C. R. Huang, and S. I. Liu, “Validity of a Novel Touch Screen Tablet-Based Assessment for Mild Cognitive Impairment and Probable AD in Older Adults,” Assessment, Dec. 2017. [85] N. Dethlefs, M. Milders, H. Cuayáhuitl, T. Al-Salkini, and L. Douglas, “A natural language-based presentation of cognitive stimulation to people with dementia in assistive technology: A pilot study.,” Inform. Health Soc. Care, vol. 42, no. 4, pp. 349–360, Dec. 2017. [86] E. Bayen, J. Jacquemot, G. Netscher, P. Agrawal, L. Tabb Noyce, and A. Bayen, “Reduction in Fall Rate in Dementia Managed Care Through Video Incident Review: Pilot Study,” J. Med. Internet Res., vol. 19, no. 10, p. e339, Oct. 2017. [87] Y. K. Ma, Y. M. Wei, S. Zhang, and W. J. Li, “Design and implementation of the smart home App based on the android system,” in Applied Mechanics and Materials, 2014, vol. 568–570, pp. 1528–1533. [88] A. Mihailidis, J. N. Boger, T. Craig, and J. Hoey, “The COACH prompting system to assist older adults with dementia through handwashing: an efficacy study.,” BioMedCentral Geriatr., vol. 8, no. 1, pp. 1–18, 2008. [89] P. Joddrell and A. J. Astell, “Implementing Accessibility Settings in Touchscreen Apps for People Living with Dementia.,” Gerontology, vol. 65, no. 5, pp. 560–570, 2019. [90] H. Eraslan Boz et al., “A new tool to assess amnestic mild cognitive impairment in Turkish older adults: virtual supermarket (VSM).,” Neuropsychol. Dev. Cogn. B. Aging. Neuropsychol. Cogn., pp. 1–15, Sep. 2019. [91] F. Ehrler and C. Lovis, “Supporting elderly homecare with smartwatches: advantages and drawbacks.,” Stud. Health Technol. Inform., vol. 205, pp. 667–671, 2014. [92] A. K. Rao, “Wearable Sensor Technology to Measure Physical Activity (PA) in the Elderly,” Curr. Geriatr. Reports, vol. 8, no. 1, pp. 55–66, 2019. [93] J. R. Thorpe, K. V. H. Rønn-Andersen, P. Bień, A. G. Özkil, B. H. Forchhammer, and A. M. Maier, “Pervasive assistive technology for people with dementia: a UCD case,” Healthc. Technol. Lett., vol. 3, no. 4, pp. 297–302, Nov. 2016. [94] V. Leuty, J. Boger, L. Young, J. Hoey, and A. Mihailidis, “Engaging older

20 D4.1 Device Selection

adults with dementia in creative occupations using artificially intelligent assistive technology.,” Assist. Technol., vol. 25, no. 2, pp. 72–79, 2013. [95] H. Aloulou et al., “Deployment of assistive living technology in a nursing home environment: methods and lessons learned.,” BMC Med. Inform. Decis. Mak., vol. 13, no. 42, pp. 1–17, 2013. [96] A. Weiss et al., “HHS Public Access,” pp. 224–229, 2020. [97] S. Bose, “Creation of an Assisted Living Environment for Elderly People using Ubiquitous Networking Technologies,” 2013. [98] Z. Hao Kevin CHONG Yu Xuan TEE Ling Jing TOH Shi Jia PHANG Jie Ying LIEW et al., “Predicting potential Alzheimer medical condition in elderly using IOT sensors - Case study,” 2017. [99] J. Sharma and S. Kaur, “Gerontechnology@_ The study of alzheimer disease using cloud computing,” 2017 Int. Conf. Energy, Commun. Data Anal. Soft Comput., pp. 3726–3733, 2017. [100] S. B. Khojasteh, J. R. Villar, C. Chira, V. M. González, and E. de la Cal, “Improving Fall Detection Using an On-Wrist Wearable Accelerometer,” Sensors (Basel)., vol. 18, no. 5, p. 1350, Apr. 2018. [101] D. L. Algase, E. R. A. Beattie, S. A. Leitsch, and C. A. Beel-Bates, “Biomechanical activity devices to index wandering behavior in dementia,” Am. J. Alzheimers. Dis. Other Demen., vol. 18, no. 2, pp. 85–92, 2003. [102] N. Jelcic et al., “Feasibility and efficacy of cognitive telerehabilitation in early Alzheimer’s disease: a pilot study.,” Clin. Interv. Aging, vol. 9, pp. 1605–1611, 2014. [103] R. Li, B. Lu, and K. D. McDonald-Maier, “Cognitive assisted living ambient system: a survey,” Digit. Commun. Networks, vol. 1, no. 4, pp. 229–252, Nov. 2015. [104] R. Al-Shaqi, M. Mourshed, and Y. Rezgui, “Progress in ambient assisted systems for independent living by the elderly.,” Springerplus, vol. 5, p. 624, 2016. [105] A. Salih, M. Salih, and A. Abraham, “A Review of Ambient Intelligence Assisted Healthcare Monitoring,” 2013. [106] J. Lapointe, B. Bouchard, J. Bouchard, A. Potvin, and A. Bouzouane, “Smart homes for people with Alzheimer’s disease: Adapting prompting strategies to the patient’s cognitive profile,” in ACM International Conference Proceeding Series, 2012. [107] K. Jekel, M. Damian, H. Storf, L. Hausner, and L. Fr??lich, “Development of a Proxy-Free Objective Assessment Tool of Instrumental Activities of Daily Living in Mild Cognitive Impairment Using Smart Home Technologies,” J. Alzheimer’s Dis., vol. 52, no. 2, pp. 509–517, 2016. [108] T. Yamazaki, “The Ubiquitous Home,” 2007. [109] D. Wilson and C. Atkeson, “Simultaneous Tracking & Activity Recognition (STAR) Using Many Anonymous, Binary Sensors.” [110] C. R. Baker et al., “Wireless Sensor Networks for Home Health Care,” 2007. [111] N. Noury and T. Hadidi, “Computer simulation of the activity of the elderly person living independently in a Health Smart Home,” Comput. Methods Programs Biomed., vol. 108, no. 3, pp. 1216–1228, Dec. 2012. [112] I. Lazarou et al., “A Novel and Intelligent Home Monitoring System for Care Support of Elders with Cognitive Impairment.,” J. Alzheimers. Dis., vol. 54, no. 4, pp. 1561–1591, Oct. 2016. [113] I. Lazarou, T. G. Stavropoulos, G. Meditskos, S. Andreadis, I. (Yiannis) Kompatsiaris, and M. Tsolaki, “Long-Term Impact of Intelligent Monitoring Technology on People with Cognitive Impairment: An Observational Study,” J. Alzheimer’s Dis., vol. 70, no. 3, pp. 757–792, 2019. [114] L. Rogerson, J. Burr, and S. Tyson, “The feasibility and acceptability of smart home technology using the Howz system for people with stroke.,” Disabil. Rehabil. Assist. Technol., pp. 1–5, Jan. 2019. [115] E. L. Tonkin et al., “Talk, Text, Tag? Understanding Self-Annotation of Smart Home Data from a User’s Perspective,” 2018. [116] Q. Liu, X. Yang, and L. Deng, “An IBeacon-Based Location System for Smart Home Control,” 2018. [117] T. Wray, P. A. Chan, E. Simpanen, and D. Operario, “eTEST: Developing a Smart Home HIV Testing Kit that Enables Active, Real-Time Follow-Up and Referral After Testing.,” JMIR mHealth uHealth, vol. 5, no. 5, p. e62, May 2017.

21 D4.1 Device Selection

Appendix

In this section, we present the tables that demonstrate the actual device presentation, along with specific characteristics. There is an initial brainstorming from the partners, then we filter our choices and a listed selection comes up for each section of devices or apps, and in the end there is always a table for the final selection of each Tier.

1. A – Tier 1: Wearable Market Search

Raw WP2 Wearable Sensor data data Comments Domain - Partner Devices types access Concept

o Reliability o Acceleration issues in Empatica E4 o Heart rate o Changes RADAR- (PPG) in CNS o Blood circadian studies HYVE, Yes volume rhythm o Difficult to CERTH pulse (PPG) o Gait contact o Electroderm speed Empatica al activity o Costs (EDA) 1690€ o Temperature

o Acceleration o Heart rate variability (PPG) o Blood volume o Changes Biovotion pulse (PPG) in HYVE, o Electro- Worn on the circadian Yes NOVAR dermal upper arm rhythm TIS Activity o Gait (GSR) speed o Temperature o Respiration rate o Oxygen saturation

Bittium eFaros o Changes o Raw in acceleration Patches on circadian Yes o Heart rate chest, device HYVE rhythm (ECG) in pocket o Gait o Temperature speed

o Changes in Fitbit o Activity / o Wrist circadian Charge HR number of worn rhythm 3 steps o Getting o Sleep o Heart rate access to quality at HYVE, No aggregates more night CERTH (PPG) granular o Daytime (minute intraday sleepines

epochs) can take s o Sleep months o Gait pattern speed

o Εlastic strap on DynaPort the lower McRoberts back Yes Acceleration o One Gait HYVE week of data storage o Wrist o Acceleration worn o 6 workout Xiaomi Mi o Heart rate modes Band 4 monitoring o Battery o Goal o Steps & No lasts up setting CERTH Distance to 20 o Different o Calories days ways to o Sleep (light wear and deep) o 40 € cost

22 D4.1 Device Selection

o Sleep o 6-axis monitorin acceleration Huawei g sensor Band 3 Pro o Sport o PPG tracking Yes cardiotacho 63 € cost CERTH and meter coaching o Infrared o Swim wear sensor monitorin o GPS g

o Measures all 4 Samsung o Acceleromet o Costs stages of Galaxy Fit er 100 € sleep No o Gyroscope o NFC o Stress CERTH o HRM Wireless measure o Heart rate Charging & guided breathing exercises

o ANT+ sensor o Battery o Acceleromet Garmin up to 9 er Forerunner days - o Heart rate 35 watch o GPS Yes Costs 140 € mode CERTH o Step count o Up to 13 o Calories h - GPS o Intensity training minutes mode o Move reminder o GPS+GLON ASS dual mode positioning o Fully o Cost 100 o PPG heart waterproo Xiaomi € rate sensor f Amazfit BIP o Suitable o Triaxial o 45 days No for CERTH acceleration of daily Running, sensor use walking, o Geomagneti (brightnes cycling c sensor s to 10%) o Barometric pressure sensor o GPS, Glonass, Galileo o Wrist heart rate monitor o Cost 200 Garmin o Option for € Forerunner an external o Music 45 heart rate o Battery: Yes Support CERTH monitor/ 13 h o Water cycling resistant device o No route o Acceleromet er o Gyroscope – Cadence tracker

o 50m Samsung water o Acceleromet Gear Fit 2 o Cost 170 resistanc er Pro € e No o Barometer CERTH o Music o Typical o Gyro Sensor Support usage o HR Sensor time: 3-4

days

o VO2 max Garmin o Acceleromet Vivosmart 4 o Cost 150 o Typical er € usage Yes o Ambient CERTH o Waterpro time: 7 light sensor of days o Barometer o Heart rate

o Battery Axivity AX3 life: 28 Raw 3D days Yes Acceleromet Costs 120 € OXF o Waterpro er of

o Light

23 D4.1 Device Selection

o Develope r friendly o Wireless SDKs & data o 3D APIs transfer acceleromet o MATLAB o IP64 Physilog er functions water and (GaitUp) o 3D Yes and APIs, dust VUMC gyroscope all resistant Barometric available o MicroUS pressure for free B for sensors o Gait and rapid file running transfer. analysis. Table 4 – Suggestions for wearable devices in the market.

1. B – Tier 1: Wearables Listed Market Selection

Raw WP2 Wearable Sensor data data Comments Domain - Partner Devices types access Concept

o Reliability o Acceleration o Changes Empatica E4 issues in o Heart rate in RADAR- (PPG) circadian CNS o Blood rhythm studies HYVE, Yes volume o Difficult to CERTH pulse (PPG) contact o Electroderm Empatica o Gait al activity o Costs speed (EDA) 1690€ o Temperature

o Acceleration o Heart rate (PPG) o Heart rate variability (PPG) o Changes o Blood Biovotion in volume circadian pulse (PPG) Worn on the Yes rhythm HYVE o Electro- upper arm

dermal o Gait Activity speed (GSR) o Temperature o Respiration rate o Oxygen saturation

Bittium o Changes eFaros o Raw in acceleration Patches on circadian Yes o Heart rate chest, device HYVE rhythm (ECG) in pocket o Gait o Temperature speed

o Changes Fitbit in Charge HR o Activity / o Wrist circadian 3 number of worn rhythm steps o Getting o Sleep o Heart rate access to quality at HYVE, No aggregates more night CERTH (PPG) granular o Daytime (minute intraday sleepines epochs) can take s o Sleep months o Gait pattern speed

o Εlastic strap on DynaPort the lower McRoberts back Yes Acceleration o One Gait HYVE week of data storage

24 D4.1 Device Selection

o Battery Axivity AX3 life: 28 Raw 3D days Yes Costs 120 € OXF Accelerometer o Waterpro of

o Light

o Develope r friendly o Wireless SDKs & data o 3D APIs transfer acceleromet Physilog o MATLAB o IP64 er (GaitUp) functions water and o 3D Yes and APIs, dust VUMC gyroscope all resistant o Barometric available o Micro- pressure for free USB for sensors o Gait and rapid file running transfer. analysis.

Table 5 – Listed selection for wearable devices in the market

1. C – Tier 1: Apps Market Search Active / Sensor data Connection to App Comments Passive Partner types Domains (WP2) App Can be used Phone o Mobile RADAR- to retrieve the interaction state phone use base total time the (Lock, unlock, Passive o Changes in CERTH app phone was startup, circadian used per time shutdown) rhythm period Metrics like o Foreground / number of RADAR- background apps used and base events total app Mobile phone Passive CERTH app o App usage time per use category is category can collected be derived from this data. o Call log Metrics like o SMS log number of RADAR- o Contact list different base contacted Social o Bluetooth Passive CERTH app devices in persons in interaction close certain time proximity can be derived. Location data can be used to derive o Relative information RADAR- location about the base o Navigation (GPS, cell mobility of Passive CERTH app o Gait speed tower, Wi-Fi) participants

o Acceleration and determine commonly visited locations.

Eye motor Cognitive Neurotr tracking while NOVA Eye movements Active function and ack looking at RTIS memory images

o Computer Software o Μeasures o Detection inconsistenc of subtle y in typing changes in Neuram Keyboard usage NOVA cadence cognitive Passive etrix (ms) RTIS o Calculation and motor of a digital function bio-marker o App download in a PC

25 D4.1 Device Selection

Advanci Driving Difficulties NOVA Active Driving ability ence simulator driving RTIS

o Either picture description Voice and Winterli task is NOVA language Active Language ght recorded RTIS biomarkers o Recordings of free speech

Active app offering tests of free reading Sonde Voice NOVA or answering Active Voice Health biomarkers RTIS questions while recording voice

Cambrid ge Cognitio n – Voice and Voice and NOVA CANTA language language Active Language RTIS B biomarkers analysis Mobile

Mind Passive data Strong collection from smartphone Interaction with Mobile phone NOVA Passive sensors mobile phone use RTIS (specifically touch screen)

o Android BeHapp Collects data of only communicative o No Mobile phone NOVA Passive or exploration interaction usage RTIS acts. with the app

Computer Simulation of Neuroco Software - preparing a g Trials Time to finish VRFCAT – meal and going NOVA App task, errors, prepare meal Active to the grocery RTIS (Navigat repetitions etc. & go with a store and buying ion) bus to buy ingredients groceries An iOS and o iPad with Android based pre- Augmented configured Altoida Reality App and Technology Web which tests the Active / NOVA Dashboard Tablet usage functional & Passive RTIS Login cognitive o Spatial aptitude of a memory, patient, via a executive self-learning functions (ML) algorithm

Dementi Response time, Reminders, Response to Active CERTH a Clock task execution daily outlook reminders

26 D4.1 Device Selection

Response time, Simulated Banking (in)correct Handling ATM Active CERTH App amounts, finances transactions passwords etc.

o Functional ability Recipes o Perceived and Task Functional Step-by- Response time, completion / Active ability CERTH step duration (combined guides with smart home devices)

Mezurio Short, daily App episodic memory tasks Input modalities in addition to including voice, Memory optional Active OXF movement and functions executive touch function and language tests.

Tweri app Not compatible with new Uses GPS to versions of iOS - find Alzheimer Alzheimer Passive CERTH requires many patients improvements of interfaces

o Mental Stimulation – FINGER Study Scientifically MindMat o Nutrition: proven to Alzheimer – e MIND diet - improve brain Cognitive Active CERTH 7.5 years health - Games impairments younger in & Entertainment terms of brain health. o Exercise o Storing Memories

o 24-hr Helpline CaringK o Individual ind and family CaringKind counseling (Formerly sessions Alzheimer’s Alzheimer - o A vast Active CERTH Association, Dementia network of New York City support Chapter). groups o Education seminars and training programs.

o Learning o Caregiving Balance o Pill Box Easily organizes App o Schedule information and Alzheimer Active o Doctor Diary CERTH serves as a o News resource. o Family

o 24-hour Helpline

Table 6 – Suggestions for apps in the market

1. D – Tier 1: Apps Listed Market Selection Active / Connection Sensor data App Comments Passive to Domains Partner types App (WP2)

27 D4.1 Device Selection

Short, daily Input episodic modalities memory tasks in including addition to Memory Mezurio Active OXF voice, optional tasks movement and executive touch function and language tests. o iPad with pre- configured App and Augmented Web Active / Memory Altoida Reality Dashboard NOVARTIS Passive tasks Framework Login o Spatial memory, executive functions Response time, Banking Simulated ATM Handling (in)correct Active CERTH APP transactions finances amounts, passwords etc. o Computer Software o Detecting functionally meaningful improvemen Time to finish ts in patients’ task, errors, everyday Activities of VRFCAT Active NOVARTIS repetitions, lives daily living etc. o Scientifically validated o Integrated data capture o 24/7 technical support Table 7 – Listed selection for apps in the market

1. E – Tier 1: Wearable Cameras Market Search

Wearable Positives Negatives Product Info Price Partner Cameras o Recording quality: 720p 30 o Unavailabl or 60 fps, 1080p e SDK for 30 fps real-time o Encoding and file transfer formatting: to an H.264 MPEG-4 external AVC .MOV file device type QuickTime GecoCam o Unavailabl o Memory: e in the through Micro market. Small size SD card, up to 187 € CERTH o Micro USB 32GB connection o Battery life: required 45Min @ 1080P for file 1Hr @ 720P transfer. o Battery: 300 o User mAh Lithium wears polymer glasses. o Connectivity: o No USB, Micro processor USB (Windows & OSX) o Low o Sensor: 5 MP battery life (file type .JPG), o No 74 ° diagonal processor optical field, 60 o SDK to ° horizontal transfer optical field, the files to PogoCam 720p HD video Small and an external in 30 FPS with 87 € CERTH discreet device in sound (file type real-time .AVI) unavailabl o Size: 1,7 x 5 e. inches o The user o Weight: 56 gr needs to o Battery: 20 days wear in standby glasses.

28 D4.1 Device Selection

o Charging time: 35 minutes to 100% o Storage: 100 photos or 30 videos o Connectivity: Wirelessly via Bluetooth o Camera: 5M pixel HD, compressed into 1,2MB o Sensor: Eye- movement – differentiates o Small and o Pre-order deliberate/usual discreet only eye-blinks. o Memory: 8 GB o SDK o No (internal) unavailabl processor o Battery: Li-on Blincam e o No real- rechargeable, o Autonomy time file 1.5 h continuous 219 € CERTH transfer use. o Easy support. o Bluetooth (BLE photo + BT2.0), Micro- capturing o The user USB via eye- needs to o IOS blinks. wear smartphones: glasses. ver. 8 or newer, Android: ver. 4.4 or newer. o Weight: 25g o Size: D x H x W: 92,4 x 17,6 x 11,6 mm. o Video analysis: 4K60, 2.7K120, 1440p120, 1080p240 o 12 MP o Sensor type: o Small and CMOS light o Optical zoom o Video recording o Attached / image refresh to many rate: 30-240fps o Video types: body parts o Does not MP4: MPEG-4/ support HVEC o Real-time voice o Microphone transfer of control in o 2 inch rotating GoPro photos Greek touch screen Hero 7 and videos language o Face 449 € CERTH Recognition o Stable o Difficult to o Additional operating attach onto microphone system a pair of input glasses o Voice o micro HDMI, control USBC port, Frame for o Long accessories autonomy connections o Wi-Fi, WLAN, o Resistant Bluetooth, GPS o Water-resistant o Weight: 117 g o Battery: Rechargeable Li-on, 2h autonomy o Camera: 5M pixel o 12h o Sensor: Vicon continuous accelerometer, Autograp recording color sensor, her o Supplied magnetometer, software PIR, (images to Temperature, GPS video). Price 280 € CERTH o Memory: 8 GB o Attempts (internal) informed o OLED display decisions o Connectivity: for best Bluetooth (BLE time to + BT2.0) Figure 12 – Vicontake o Weight: 58g pictures. o Width 37.4mm (with side

29 D4.1 Device Selection

buttons); length 90mm (95.5mm with lanyard ring); thickness 22.9mm (with clip and lens). o Image quality Snapchat 1216 x 1216 Spectacle (videos) and s 2 o Can 1642 x 1642 record (pictures) 150 € CERTH under o Can hold up to water 150 videos or 3000 photos at a time o Egg-shaped charging case that recharges up to two cameras. o Dust and water resistant. o 1080x1080 @ OPKIXON o Weights 30 fps E just 11 gr o Camera Each camera’s battery lasts housing: 300 € CERTH o Charging only for 12 aerospace case: 6 minutes aluminum camera o 4 GB storage – charges 15 mins of video o Egg storage: up to 70 mins of capture o Easy-to-use OPKIX Studio App for editing o Share button o 130 degrees o Unflatterin max-view angle g, poorly o 16 GB flash framed internal storage videos o Battery life: Google o Not water Video recording Clips resistant for 3 hours Does not o 15 fps frame- 170 € CERTH violate privacy o Needs a rate case that o Wi-Fi Direct sticks to a connectivity, wall, or remote control one with a via mobile tilting devices, remote aspect live view o Hi-Res 12 MP Images o Full HD 1080p Video o 1.8 inch LCD o Ultra-Light Display SereneLif Weight o Weather- e Clip-on o Very high- Resistant Wearable o Built-in WiFi Camera quality images o SereneViewer 50 € CERTH and videos App to monitor the location o Free app: o Micro SD-card “Snap slot for storage Pics” o Night Vision LED Illumination Lights o Weight: 1.85 lbs.

Table 8 – Suggestions for wearable cameras in the market

1. F – Tier 1: Wearable Sensors Categorization Wearable Sensors Sensor Measurement Data Rate Accelerometer Acceleration High Gyroscope Orientation High Glucometer Blood Glucose High Pressure Blood Pressure Low

CO2 Gas Respiration Very low ECG (Electrocardiography) Cardiac Activity High

30 D4.1 Device Selection

EEG (Electroencephalography) Brain Activity High EMG (Electromyography) Muscle Activity Very high EOG (Electrooculography) Eye movement Very high Pulse Oximeter Blood Oxygen Saturation Low GSR (Galvanic Skin Response) Perspiration Very low Thermal Body temperature Very low Table 9 – Wearable Sensors Categorization

1. G – Tier 1: Wearable Literature Search The IoT technology aspect considers the various IoT wearable sensors and devices found in earlier review studies, mainly categorized in “wearables”, “smartphones”, “robotics”, “smart home”, “environmental sensors”, “indoor positioning”, “biometric sensors”, (fixed) “cameras”, “wearable cameras”, “microphone”, and “applications”. While all categories refer to specific hardware, the latter refers to any type of software and AI algorithm on local PCs or the cloud, which does not require a hardware IoT component of the former categories. The study in [26] considers five types of devices: PIR motion sensors, body-worn sensors, pressure sensors, video monitoring, and sound recognition. Our review generalizes further to include more device types that are not considered there; for example, PIR motion sensors and pressure sensors are included in “smart home” sensors along with other possible types such as door-widow sensors, appliance and object usage sensors, and so on. Body-worn sensors are essentially “wearables”, and video monitoring and sound recognition are mapped to “cameras” and “microphones”, respectively, in our review. To begin with, “wearables” are dominant in the literature, owing to their increasing popularity and affordability. Cedillo et al. [27] selected the most relevant devices to an AAL context, combining “wearables” and “applications” that contribute to the wellbeing of elders. Piwek et al. [28] includes various types of wearables, such as headbands, sociometric badges, camera clips, smartwatches, and sensors embedded in clothing, while Haghi et al. [29] deal with nine different motion trackers and four commercially available wrist-worn devices in the market for vital signs measurement, that is, FitBit, Jawbone, Withings, and Misfit. Another study [30] complements this list of commercial wrist-worn devices with Apple iWatch, Samsung Gear S2, Pebble Time, UP4 by Jawbone, Empatica, and Fitbit Flex, among others, through head-mounted devices and other accessories, such as smart jewelry, e-textiles, skin patches, and even an e-tattoo. In addition to both commercial devices and research prototypes, this review also examines pertaining potential security threats and confidentiality issues. Surendran et al. [31] explores smart wearable locator band, smart socks, the CleverCare Smart watch, iTraq, MedicAlert Safely Home, PocketFinder, Trax, and wearable cameras. Biometric sensors are a special type of wearable or non-wearable devices that are used for both continuous and on-demand measurement of physiological and medical data. While they are often applied to security, for example, through fingerprint scanning, they are also used in healthcare, for example, measuring body temperature, electrocardiogram (ECG), pulse oxygen saturation, blood pressure, blood glucose, and so on [32]. Patel et al. [33] examines both smart home sensors for in-house positioning and microphones to record audio and voice, as well as a wide range of biometric sensors for glucose, pH, and O2 measurements. Another study [34] deals with what IoT offers to the neurological aspects of health disorders, examining devices that can be classified as both “wearables” and ”biometric sensors”, such as the Basis Health Tracker, Misfit Shine, Fitbit Flex, Withings Pulse O2, Actiwatch Spectrum, FitBit, Empatica 4, Bittium Faros, and PhysioCam. It also mentions an in-ear sensor for EEG (electroencephalogram). The study in [35] examines four wearables from a medical point of view, namely, Myo, Zyo patch, MyDario, and SleepBot. Along those lines, the study of [36] examines wearables and biometric sensors for diabetes, heart monitoring, and pulmonary disease, including radio-frequency identification (RFID) and wireless sensor networks (WSN) parameters.

Health Reference Wearables Worn on Aim Partner Focus

Khojasteh Fall If fall is detected, a Smartwatch Arm CERTH et al. [37] Detection series is done.

SmartMOV E app, iPod Footswitch Accuracy of Touch, Fall on foot, smartphone-based Ellis et al. footswitch Detection, GAITRite gait analysis via CERTH [38] sensor, Parkinson on the two measurement GAITRite head devices. sensor.

31 D4.1 Device Selection

Attached Differences in to a Velcro Small, walking and turning Weiss et elastic belt lightweight Parkinson between PD CERTH al. [39] on their sensor. patients and lower healthy subjects. back.

A single wearable Mc Ardle et Acceptability and (tri-axial Alzheimer Arm CERTH al. [40] feasibility of gait. acceleromet er)

Sensors attached MPU6000 to five (triaxial body Postural tasks with Costa et acceleromet Alzheimer segments: increase in CERTH al. [41] er & trunk, both difficulty. gyroscope) legs and both thighs.

A wearable sensor (triaxial acceleromet er, gyroscope, Identification of Zhou et al. and Attached Alzheimer motor cognitive CERTH [42] magnetome to the shin impairment. ter) combined with a human- machine interface.

Inertial- sensor- based Mounted wearable on Walking test (single Hsu et device (a Alzheimer participant and dual task) and CERTH al. [43] triaxial s’ feet & balance ability test. acceleromet waist. er and two gyroscopes)

2 body Shimmer sensors (waist) (waist & Alzheimer, A fall monitoring Abbate et and head), in- Fall system in a long- CERTH al. [44] Enobio house Detection term nursing home. (head) sensors and sensors. cameras.

A tracking device combining GPS & Pot et al. Patient indoor General Alzheimer Belt-worn CERTH [45] tracking. Packet Radio Service (GPRS)

o Accuracy of the Around StepWatch Stroke, the right o Investigate PD, MS, ankle Giggins et relationships NOVAR StepWatch Stroke, proximal al. [46] between daily TIS TBI, Rett to the step counts, Syndrome lateral gross motor malleolus skills, and age.

o Accuracy in The wrist identifying Giggins et Nike Stroke, of the less NOVAR stepping al. [46] Fuelband TBI affected TIS activity in arm people with stroke

32 D4.1 Device Selection

o Comparison with StepWatch

o Effect of walking speed on the Each accuracy of an participant accelerometer- ’s non- based activity Giggins et paretic monitor after NOVAR FitBit One Stroke al. [46] side on a stroke. TIS waistband o Effect of and ankle position (waist strap vs. ankle) on the accuracy of an activity monitor

Mounted onto waistband s and fitted Actical Giggins et MS, around the NOVAR acceleromet Activity count al. [46] Stroke participant TIS er s’ waists over the iliac crest of the left hip

Doppler- Recovery, Continuous Lie et al. based vital prevention Monitoring of Vital- OXF [47] signs or Signs biosensor prediction

Investigate the feasibility of a Portable Near the portable preimpact Preimpact body fall detector in Wu and Fall Fall center of detecting OXF Xue [48] Detector Detection gravity, impending falls With Inertial inflatable before the body Sensors hip pad impacts on the ground.

Several Several triaxial triaxial Stats for acceleration sensor Lai et al. Fall acceleration each devices for joint OXF [49] Detection sensor position sensing of injured devices body parts.

Analyze the discriminative power of A wearable electrodermal device was activity (EDA) in used to distinguishing Setz et al. monitor the Stress Forearm stress from OXF [50] electroderm cognitive load in an al activity office environment (EDA) – keeping track of stressful phases during a working day.

Table 10 – Suggestions for wearable devices in the literature

1. H – Tier 1: Apps Literature Search

Health Reference App Features / Aim Partner Focus

o Provides caregivers with service and Aljehani iCare Alzheimer updates of patients CERTH et al. [51] o Provides patients with more freedom

33 D4.1 Device Selection

o No internet connectivity required. o Access to free ALZ Alzheimer’s and other Caretaker dementias caregiver App, part of resources and training Pirani et the Alzheimer materials. CERTH al. [52] Alzheimer’s o 24-hour care giving companion assistance via toll free system. phone number or email submission. o GPS tracking unavailable.

o Runs on the smart watch. o Collects the Karakaya Android accelerometer and Elders CERTH et al. [53] mobile app gyroscope readings. o Uploads them periodically to an application server.

o Utilizes the smartphone’s inertial SmartMOVE measurement unit to Fall Ellis et al. (on an record gait movements Detection, CERTH [38] Apple iPod during walking. Parkinson Touch) o Enables precise control over testing parameters.

Table 11 – Suggestions for apps in the literature

1. I – Tier 1: Wearable Cameras Literature Search

Referenc Wearable Health Aim Partner e Devices Focus

Cognitive training groups A. Silva et SenseCam Alzheimer for cognitive state CERTH al. [54] improvement.

SenseCam E. (temperature, External memory aid to Woodberr light and a Alzheimer promote recall of CERTH y et al. passive infrared episodic memories. [55] sensor) and a laptop.

Event segmentation to check to which extent possible impairments in Allé et al. A wearable Schizophreni temporal ordering and CERTH [56] camera a segmenting in patients hinder memories construction.

Whether video clips captured from a body- worn camera every time FMT: Fiducial objects of interest are Marker Tracker. found within its field of F. M. Li et A wearable Elders view can help older CERTH al. [57] camera-based adults determine if they object have completed certain actions with these objects and what their states are.

We suggest that factors relating to rehearsal / reconsolidation may SenseCam aiding Berry et Limbic have enabled SenseCam autobiographical CERTH al. [58] encephalitis images to improve the memory subject’s autobiographical recollection.

34 D4.1 Device Selection

Periodic review of Severe images of events Hodges SenseCam memory recorded by SenseCam CERTH et al. [59] impairment resulted in significant recall of those events.

SenseCam supported rehearsal of memories of events that trigger high levels of anxiety would yield improved retrieval of both factual detail and Acquired Brindley internal state information SenseCam Brain Injury CERTH et al. [60] (thoughts and feelings) (ABI) compared with a conventional psychotherapy aid (automatic thought record sheets, ATRs) and no strategy.

Table 12 – Suggestions for wearable cameras in the literature

2. A – Tier 2 & 3: Smart Home Devices Market Search Data Smart Access Sensor data Device WP2 Domain Home (Strea types Partner Characteristics – Concept Device m or (Modalities) Log) ADLs e.g. Cooking, Plugwise On-off, works o Affordable, house Circle any power reliable and chores, (Plug) supplied fast (40€ per watching TV Stream appliance, CERTH appliance) – combined even non- o Useful for with socket ADLs presence appliances sensors and processing o Affordable, Plugwise reliable and Scan fast (Motion) Presence in ADLs e.g. o Integrates Stream a room (IR bathroom CERTH with plugs motion) visits easily o Useful for ADLs o Affordable Special (60 € per events e.g.: room) o Too bright o Reliable and Plugwise outside Environment fast Sense and lights al o Integrates Stream on CERTH (temperature, with plugs o Too luminance) easily cold/hot o Useful for and A/C special or heater event on, etc. detection o Very CAO Affordable (20€ per Gadgets Object ADLs – more object) Tag movement accuracy for o Not so Sensor Stream (Only known e.g. cooking, CERTH reliable sensor that fiddling with (delay, hard does that) meds, chores to setup) o Useful for ADLs At-home or taking a walk, o Very doing outside Various Affordable Open door / chores (after Door Stream o Useful for CERTH window processing, Sensors presence at combined home with other sensors)

35 D4.1 Device Selection

o Person’s proximity to certain zones (cooking, o Affordable watching (99$ for 4 TV, Estimotes, phone, Estimote 110€ for 2 Object / bathroom, beacon Stream Beacons) Person outside CERTH / Log proximity etc.) o Useful for o Special special events event (forgetting detection a purse, keys, mobile phone etc.) o Sleep stages & interruptio Sleep quality Beddit or ns and problems Nokia o Time to after o Affordable Sleep fall processing (60€) Sensor asleep (too many CERTH, Log o Very reliable etc. (insomnia interruptions, JANBE and / short/long unobtrusive restlessn duration ess) according to o Total time thresholds) asleep / per stage DREEM Portable 599 € (unlimited EEG metrics Log EEG for access to raw and Sleep JANBE sleep data) Quality

o Portable EEG o EEG Muse o Performa metrics (mini nce / o Cost 200€ (after EEG) Log score in o App guides processin CERTH CBT self-setup g) o Raw Data o CBT (with iOS scores app?)

o Activity Emerald; o Cost level Monitor WiTrack unknown o Sleep motion https://ww o Very monitorin throughout w.emerald unobtrusive g home via inno.com/ Stream o Inability to o Gait NOVARTIS changes in identify the paramete radio / http://witra person rs wireless ck.csail.mi o Expensive o Fall fields t.edu detection o ADL o Cost unknown o Very o Monitor Monitor unobtrusive respiratio Origin motion o Early-stage n Wireless through technology o Sleep http://origi Stream TAKEDA home via under co- monitorin nwirelessa changes in developed g i.com Wi-Fi signal with many o Activity hardware detection vendors

o More o EEG annoying / sensors – obtrusive for 14 the user (VS o Monitor EMOTIV channels Dreem) brain EPOC+ Log o Include o The user activity CERTH (mobile Motion must be o EEG EEG) sensors awake metrics o Raw EEG o More EEG available sensors than Dreem GARMIN o Cost 60€ Heart rate HRM Stream Portable HR o App guides CERTH metrics Premium self-setup

36 D4.1 Device Selection

MEDISAN A Sleepace Sleep SC 800 Portable Log Costs 130 € parameters CERTH sleep sensor metrics

PlayBrus h Portable Tooth brush - Costs 15 € CERTH Tooth brush monitoring

GOCLEV ER Car DVR Titanium Driving, GPS Log Car camera Cost 50 € CERTH GPS camera

Smart Fork SLOW CONTRO Smart Fork, Eating L 10S - Cost 30 € Eating, ADLs CERTH measuremen t

MEDISAN A TM735 Thermom eter Smart Temperature Stream Cost 50 € CERTH Thermometer levels

Bluetooth Tracker LAPA 2 Finding Object Stream Cost 20 € objects CERTH detection (keys)

ADLs e.g. Cooking, TP-LINK o Affordable, house HS100 reliable and chores, Smart Electric fast (30€ per watching TV Plug Stream Appliance appliance) CERTH – combined on/off o Useful for with ADLs presence

sensors and processing o Occupancy sensor o Window Modular contacts security Gateway- o Wall concept to EnOcean connected switches meet the system Stream o Room power CERTH (TCP/IP) temperature requirements controlled via sensor of energy smartphone o Heating harvesting valve systems o Plug-in receiver o Security concept REVOGI when Smart away Sense from o Motion home Starter Combined sensors o App Kit Smart Log o Theft included CERTH Sensor sensors o Functions Solution o Speaker even without internet connectio n o 45 €

37 D4.1 Device Selection

XIAOMI o Easy to Mi Smart o Motion plug & Sensor Sensor connect Set Combined o Control Hub o Control Smart o Window and through Stream CERTH Sensor Door Sensor your Solution o Wireless smartpho switch ne o Speaker o Safety o 130 € o Enhanced PANASO smart NIC security Smart o Push Home Pairing DECT Ultra Safety o Motion o Wide Low Energy Starter Stream sensor range CERTH wireless Kit o Door sensor (300 m) protocol o Easy to connect o App included o 80 € o On / off on each sensor o iPhone compatibl e o CubeOne iSMARTA o Shows device LARM who is in o 2 Door / Combined the house Window Smart o Call, Log sensors CERTH Sensor SMS, o Motion Solution Push sensor notificatio o 2 tags n or email o 2 stickers for informatio n of unusual activity. o 90 € o Emergency Essential button o Gateway Kit o Smart Plug and Special Combined o Door / smartwatt Care Smart Window Stream service CERTH Sensor sensor are Solution o Smart required. lightbulb o 140 € o Motion sensor

Essential o Temperature o Gateway Kit Flood sensor and Combined & Fire o Leakage smartwatt Smart Stream sensor service CERTH Sensor o Smoke are Solution sensor required. o Indoor Siren o 136 €

Cube o Motion o Reminds to Sensors Sensor go to bed o Light o Sleep Log ADL CERTH Sensor patterns o CO2 o Environment Sensor changes o Accelerati BLE on o Batteries Beacons Log o Temperat ADL CERTH o USB and Tags ure o Light o Bluetooth o Low-Energy o From inches Gimbal Motion to 50 meters Log ADL CERTH Beacons Sensors o Indoor and outdoor use o Long battery life

38 D4.1 Device Selection

OBD2 car Bluetooth GPS diagnostic Bluetooth, Stream interface Motion, Driving Ability CERTH scanner Acceleration (Android)

Table 13 – Suggestions for Smart Home Devices in the market 2. B – Tier 2 & 3: Smart Home Apps Market Search Connecti Smart Sensing or Sensor on to Home Input Features Partner types Domains App (Modalities) (WP2) Remote pre- screening and DeepSpA Cognitive Telephone/ monitoring Passive Function, JANBE video solution for Speech clinical trials in cognitive decline o All smart devices together o Personaliz ed smart home routines o Light o Adjust o Sleep sensor home o Daily o Hub automatic aYonomi activities o Switche Sleep CERTH ally to the o Managin s perfect g o Thermo setting stat etc. throughou t the day. o Home responds when you arrive or leave. o Control your smart devices with easily to create Triggs. Olisto o Smart notificatio o Thermo ns for stats location, o Lights Electric Managing actual o Other Appliance CERTH apps energy smart s consumpti devices on, etc. weather conditions , sports etc. o Voice integration – Amazon, Google o Organizes home products in rooms, zones and myHome service groups for Plus o Lightbul ADL easy b organizing: managem ADL, o Fan o Rooms ent. Electric o Door CERTH o Zones o Triggers – appliance o Lock o Service multiple s o Switch Groups actions at etc. specific time, location, or event (e.g. close all windows

39 D4.1 Device Selection

when leaving in the morning). o Multiple users o Voice control o The first platform centralizin ImperiHo g the me o Urban Smart mobility devices o Energy and the IoT platform consum services Transport for smart CERTH ption of the ation cities meters Smart o Air City. quality o Does not need any external server access o Voice o Home control Samsung devices o Hood, Home o No oven & Connect connect kitchen ion with IFTTT control home Cooking, service and o Washing CERTH function ADL “Applets” machine s control (lights, o Fridge heat, control window (cameras s etc.) inside) Table 14 – Suggestions for apps in the market

2. C – Tier 2 & 3: Ambient Sensors Categories Ambient Sensors Measureme Data Sensor Aim Installation nt format Stove use, room temperature, use of water, and PIR Motion Categorical opening of Wall, Ceilings cabinets, showering, presence in key areas at home To determine the Window, Door, Active Motion, Categorical presence of Toilet, Infrared Identification someone Bathroom To determine the Object RFID Categorical proximity between Any object Information two objects o Detect presence, respiration, Floor, pulse, furniture, bed, Pressure on movement in Pressure Numeric gas system, Mat, Chair the bed water system, o Amount of shoes, pillow water/gas used o Sitting posture analysis. o Person presence Smart Pressure on Numeric o Object removal Floor Tiles Floor (chair, table etc.) Stores a categorical Magnetic Door, Cabinet value when a Categorical Many places Switches opening / closing magnetic field affects it Utilizes HF waves to Ultrasonic / Ultrasonic Numeric Motion detect radar motions Motion, Camera Image Activity recognition Ceiling Activity Micropho Activity recognition Activity Sound Ceiling ne via sound

40 D4.1 Device Selection

Table 15 – Ambient Sensor Categorization

2. D – Tier 2 & 3: Literature Search Regarding IoT Wearable Sensors and Devices for Eldercare. Smart home devices are usually ambient and inobtrusive in an AAL context. A study from Wang [61] reviews indoor positioning systems, emphasizing on human activity recognition, as well as biometric sensors (vital sign monitoring, blood pressure, and glucose). Blackman et al. [62] consider three generations of AAL, gathering 64 studies, and consider parameters such as social support, interface, and health monitoring capabilities. They include wearables and smart home sensors (AiperCare, Aladdin, bed occupancy sensor, and so on), as well as environmental sensors such as gas detectors. The review in [17] deals with most types of “smart home” ambient sensors, “wearables” and “wearable cameras”, e- textiles, and “indoor positioning” systems, especially oriented around AAL projects. Fall detection, prevention, and risk assessment mainly involve wrist-worn sensors, RFID sensors, and a footwear, as reviewed in Baig et al.[63]. The researchers in [64] also review AAL platforms, with wearables and smart home sensors to enable multimodal fall detection. Related to that, Ienca et al. [65] cover a wide area of intelligent assistive technologies around mobility and rehabilitation aid.

Reference Sensors Aim Outcomes Partner

PIR motion sensors, stove Recognition Alwan et al. sensor, bed of activities of High acceptance OXF [66] pressure daily living. sensor.

PIR motion Ariani et al. sensors, Fall Detection 89.33% accuracy OXF [67] pressure mats.

The functionality of the system in Video Recognition detecting activities Bamis et al. monitoring, PIR of activities of and deviations in OXF [68] motion daily living. patterns of sensors. activities was described.

Doppler radar Chung et al. Sleep stage and 100% accuracy. OXF [69] classification microphone

54% improvement PIR motion compared to a Guettari et sensors and Localization standalone one OXF al. [70] sound sensors. multimodal system.

o Main advantage: ease of tracking the users. o Main Video camera, Recognition Kinney et al. disadvantage: PIR motion of activities of OXF [71] annoyance by sensors. daily living. false alerts. o $400 to equip the house, $90 per month the cost of maintenance.

o Identification of abnormal behaviour. PIR motion o Satisfactory sensors, door- Recognition Lotfi et performance opening of activities of OXF al. [72] in identifying sensors, flood daily living health status sensors. using different ambient sensors.

41 D4.1 Device Selection

Video camera, A change of PIR motion health status was Rantz et al. sensors, bed Fall detection detected by the OXF [73] pressure system, ignored sensors, door by the nurses. sensors.

PIR motion Recognition Improved the Van Hoof et sensors, video of activities of sense of safety OXF al. [74] camera. daily living. and security.

Video camera, Recognition Zhou et al. 92% precision; PIR motion of activities of OXF [75] 92% recall. sensors. daily living.

Video camera, Recognition 62–94% Zouba et al. PIR motion of activities of precision; 62– OXF [76] sensors. daily living. 87% sensitivity.

o Sound All parameters (snoring) Rachakonda Stressfulness, measured, can be o Sleep hours CERTH et al. [77] Sleep a very useful o Respiratory scientific tool. o Heart rate

Table 16 – Suggestions for IoT Wearable Sensors and Devices in the Literature

2. E – Tier 2 & 3: Apps for Dementias Literature Search

Reference Apps Aim Outcomes Partner

TV-based platform service to support Challenges such as: patients with mild Goodman- cognitive Casanova ΤV-AssistDem o Poor engagement CERTH impairment and et al. [78] o Connectivity relieve their problems. caregivers. Improve quality of life.

Ticket To Talk – prompt o Promoting and Encourage carers to managing conversation collect & reminiscence between younger Welsh et organize o Starting and people and their CERTH al. [79] media, use maintaining old relatives them as conversation living with prompts and o Redistributing dementia. conversation Agency starter.

To report that MCI patients P-values in t-test have a show that token type Shibata et VocabChecker significantly ratio of PwD is higher CERTH al. [80] larger vocabulary than CG (p<0.05, than healthy highly significant). elderly do.

PainChek – To describe a pain scale, novel system o Sound display pain focusing on its psychometric assessment conceptual properties log, pain chart foundation, o Excellent and local Atee et al. clinical and concurrent validity, patient CERTH [81] technical interrater database, contents, clinical reliability, internal provide use, and consistency medications practical tips for o Excellent test re- and therapies use in clinical test reliability and comments settings. section.

Tyack et Art Viewing Explore whether o Improvement in CERTH al. [82] App – art-based well-being presenting intervention can

42 D4.1 Device Selection

choices of art be delivered via o Enthusiasm genre to view a touchscreen from museums tablet device and area displaying art artists. images

o Positive usability Improve the pain testing Pain assessment of o Paramedics Assessment PwD (People Students and Docking App – with the with Dementia) Delphi panel of CERTH et al. [83] Abbey Pain and experts state it as Scale based management in a useful tool in a questionnaire this vulnerable pre-hospital population. setting

o Mann–Whitney U Develop a tests were psychometrically Attention, significant for CDR valid Visual Memory, = 0 versus 0.5, touchscreen Visuospatial, and CDR = 0 tablet-based Huang et Reaction Time, versus 1 cognitive test CERTH al. [84] Fine Motor o Confirmation of battery to identify Control. (App the four-factor early cognitive name not model by impairment due mentioned) Confirmatory to dementia in factor analysis older adults. (CFA).

Provide elderly Wizard-of-Oz people and PwD System with computer-based o Enjoyment Dethlefs sorting, cognitive o Wanted to use it CERTH et al. [85] naming, recall, simulation via again quiz and spoken natural proverb. language.

SafetyYou App –video view o Analyse how feature from continuous past 72 hours, video o Drop in fall rate Bayen et live video view monitoring o Positive impact on CERTH al. [86] on a o Review of quality of care. smartphone falls of PwD from each can support a installed better quality camera. of care.

o Monitors the remote data and controls the appliances in a Enhancing the long distance Ma et al. Android-based portability of the o Receives the CERTH [87] app smart home sensor data timely management o Controls the appliances properly

o Completed 11% more Guide an older handwashing adult with steps Mihailidis App (with a dementia independently CERTH et al. [88] device) through the ADL o 60% fewer using audio and / interactions with a or audio-video caregiver prompts o The accessibility features significantly o Solitaire improved usability o Enable Joddrell o Bubble in Solitaire customisation and Astell Explode o Bubble Explode CERTH o Improve [89] Both with retained the high usability accessibility level of usability settings without further improvements

aMCI group: Investigate o Lower Eraslan Virtual cognitive performance Boz et al. CERTH Supermarket functioning o Required more [90] (VSM) time to complete the VSM

43 D4.1 Device Selection

Table 17 – Suggestions for apps for elders with forms of dementia in the literature

2. F – Tier 2 & 3: Review of Case Studies of IoT Wearable Sensors and Devices for Eldercare This section presents a detailed review of case studies of IoT wearable sensors and devices for eldercare. In this paper, we use the general term “case study” to refer to any published study related to the topic, of any type. Some of them might be observational, interventional, or usability studies. These can be discriminated in this review through their “aims”, which are usually “monitoring” and “intervention” for observational and interventional studies, respectively. Usability studies can be discriminated through their evaluation outcome measures, which are mostly “acceptance”, “user satisfaction”, and “feedback”. All those aspects and categories are explained below. The Table below presents what we have identified as recent and representative works according to the previous classification. The following subsection presents how the studies consider each aspect. Studies with an evaluation experiment are examined at the end as they entail even more aspects pertaining to evaluation duration, participants (cohort), and outcome measures. Most case studies found present a device, or more, that measures different parameters of a disease. Alzheimer’s disease (AD) is the most common disease included in our study. There are some studies that focus on devices for the general elder population [91], dementia [92]–[95] Parkinson’s disease [96], and fall detection [53][37], or even combining some, or all, of the above mentioned [38], [97]. Some of the reviews presented the aim of the detection of specific symptoms or behaviors arising from a person that has a known disease. Such review studies are [98] [39] [40]. Another focus of this category is monitoring [44], [99]. These studies refer to a monitoring system for patients with a specific disease. Furthermore, another aim found in [93], [95], [100] is development. These reviews focus on developing an algorithm or a specific architecture for a system, so it measures specific characteristics. Two studies [91], [101] focus on comparing the reviewed subjects, while two others’ [45], [51] aim is tracking the patient, so the caregiver can be more comfortable or even the subject themself can be more independent. Finally, there are also aims such as biometric measurements [51], recall of some memories [55], patients’ improvement [94], and rehabilitation [102].

Reference Sensors Health Focus Aim Partner

Repetitive patterns Hao et al. In-house IoT Alzheimer detection for potential CERTH [98] Sensors Alzheimer.

Dementia, Li et al. Review / Survey, Smart Home Chronic CERTH [103] Smart Home Projects. Diseases

Review / Survey Blackman Indoor Positioning dealing with ease of Elders CERTH et al. [62] Sensors usage, safety and emergency.

Review / Survey dealing with ease of Al-Shaqi Indoor Positioning Alzheimer, usage, signal type, CERTH et al. [104] Sensors Dementia cost and efficacy of studies.

Review / Survey Alzheimer, Smart Home, dealing with Salih et Dementia, Environmental communication CERTH al. [105] Cardiovascular Sensors techniques and Diseases security.

Guidelines to help researchers maximize the efficiency of smart Lapointe homes by adapting the Smart Home Alzheimer CERTH et al. [106] form of prompts to the specific cognitive profiles of patients with AD.

Smart home, activity Jekel et Investigate the sensors, video Alzheimer CERTH al. [107] potential of a smart cameras home environment for

44 D4.1 Device Selection

the assessment of IADL in MCI.

o Video cameras o Behavior o Microphones Yamazaki General smart- monitoring o Floor pressure OXF [108] home project o Tracking personal o Motion items o RFIDs

o Video cameras o Resting hours Rantz et o Bed pressure General smart- o Behaviour OXF al. [73] o Stove door CSS home project monitoring o Motion

Wilson o Motion detectors o Resident’s location and o Pressure mats General smart- o Behaviour OXF Atkeson o CSSs home project monitoring [109] o RFIDs o Prediction

o Accelerometer o Blood pressure Baker et readings General smart- Healthcare monitoring OXF al. [110] o Microphones home project o Heart rate o Temperature

Noury and General smart- Producing elderly’s life Motion Sensors OXF Hadidi home project scenario [111]

o Propose a system for continuous and objective remote monitoring of o Wearable problematic daily sensors living activity areas o sleep sensors o Design o object motion personalized Lazarou sensors Dementia interventions CERTH et al. [112] o presence based on system sensors feedback and o utility usage clinical sensors observations for improving cognitive function and health-related quality of life

o Ambient depth Investigate the long- cameras term effects of o Plug sensors Assistive Technology o Tags Lazarou Cognitive combined with tailored o Presence CERTH et al. [113] impairment non-pharmacological sensors interventions for o A Sleep sensor people with cognitive o Wristwatch impairment. (wearable)

Table 18 – Suggestions for Case Studies of IoT Wearable Sensors and Devices for Eldercare in the literature

2. G – Tier 2 & 3: Smart Home Apps Literature Search Smart Sensing or Home Input Sensor types Features Partner App (Modalities) o Enhanced sense of security for the user and Feasibility their family that help is at Howz and o Light hand if needed. app acceptability o Temperature CERTH o If activity changes, an [114] for stroke o Movement alert is then sent to the survivors. user and the named contact. Multi- Self- modal annotation: Room-based list, 12 people encouraged to use annotati strengths voice, or NFC a multi-modal annotation app CERTH on app and tags. in a smart home. [115] limitations

45 D4.1 Device Selection

Remote pointing Android Personalized menu of remote control of app BLE Beacons “one-click” control in a CERTH various [116] smartphone app. household Appliances App uses data from light Home- sensors on beacons to MSM based self- monitor when HBST kits are app testing Beacons CERTH opened, facilitating timely [117] (HBST) for follow-up phone contact with HIV users. o Door – Adapting to window climate sensors MIT App changes in o Temperature Wind and raindrop detection CERTH Inventor the indoor sensor and outdoor o Wind sensor environment o Chemical sensor (CO2) Table 19 – Suggestions for smart home managing apps in the literature

46