DOCSLIB.ORG

Deep, Convolutional, and Recurrent Models for Human Activity Recognition Using Wearables

Total Page：16

File Type：pdf, Size：1020Kb

Download full-text PDF Read full-text
Deep, Convolutional, and Recurrent Models for Human Activity Recognition Using Wearables Proceedings of the Twenty-Fifth International Joint Conference on Artificial Intelligence (IJCAI-16) Deep, Convolutional, and Recurrent Models for Human Activity Recognition Using Wearables Nils Y. Hammerla1,2, Shane Halloran2, Thomas Plotz¨ 2 1babylon health, London, UK 2Open Lab, School of Computing Science, Newcastle University, UK [email protected] Abstract these relatively simple means suffice to obtain impressive recognition accuracies. However, more elaborate behaviours Human activity recognition (HAR) in ubiquitous which are, for example, of interest in medical applications, computing is beginning to adopt deep learning to pose a significant challenge to this manually tuned approach substitute for well-established analysis techniques [Hammerla et al., 2015]. Furthermore, some work has sug- that rely on hand-crafted feature extraction and gested that the dominant technical approach in HAR benefits classification methods. However, from these iso- from biased evaluation settings [Hammerla and Plotz,¨ 2015], lated applications of custom deep architectures it is which may explain some of the apparent inertia in the field difficult to gain an overview of their suitability for toward the adoption of deep learning techniques. problems ranging from the recognition of manipu- Deep learning has the potential to have significant impact lative gestures to the segmentation and identifica- on HAR in ubicomp. It can substitute for manually designed tion of physical activities like running or ascending feature extraction procedures which lack the robust physio- stairs. In this paper we rigorously explore deep, logical basis that benefits other fields such as speech recog- convolutional, and recurrent approaches across nition. However, for the practitioner it is difficult to select three representative datasets that contain movement the most suitable deep learning method for their application. data captured with wearable sensors. We describe Work that promotes deep learning almost always provides how to train recurrent approaches in this setting, in- only the performance of the best system, and rarely includes troduce a novel regularisation approach, and illus- details on how its seemingly optimal parameters were dis- trate how they outperform the state-of-the-art on a covered. As only a single score is reported it remains unclear large benchmark dataset. We investigate the suit- how this peak performance compares with the average during ability of each model for HAR, across thousands parameter exploration, and how likely a practitioner would be of recognition experiments with randomly sampled to discover a system that performs similarly well. model configurations, explore the impact of hy- In this paper we provide the first unbiased and systematic perparameters using the fANOVA framework, and exploration of the performance of state-of-the-art deep learn- provide guidelines for the practitioner who wants ing approaches on three different recognition problems typi- to apply deep learning in their problem setting. cal for HAR in ubicomp. The training process for deep, con- volutional, and recurrent models is described in detail, and 1 Introduction we introduce a novel approach to regularisation for recur- Deep learning represents the biggest trend in machine learn- rent networks. In more than 4,000 experiments we investigate ing over the past decade. Since the inception of the umbrella the suitability of each model for different tasks in HAR, ex- term the diversity of methods it encompasses has increased plore the impact each model’s hyper-parameters have on per- rapidly, and will continue to do so driven by the resources of formance, and conclude guidelines for the practitioner who both academic and commercial interests. Deep learning has wants to apply deep learning to their application scenario. become accessible to everyone via machine learning frame- Over the course of these experiments we discover that re- works like Torch7 [Collobert et al., 2011], and has had sig- current networks outperform the state-of-the-art and that they nificant impact on a variety of application domains [LeCun et allow novel types of real-time application of HAR through al., 2015]. sample-by-sample prediction of physical activities. One field that has yet to benefit from deep learning is Hu- man Activity Recognition (HAR) in Ubiquitous Computing 2 Deep Learning in Ubiquitous Computing (ubicomp). The dominant technical approach in HAR in- Movement data collected with body-worn sensors are multi- cludes sliding window segmentation of time-series data cap- variate time-series data with relatively high spatial and tem- tured with body-worn sensors, manually designed feature ex- poral resolution (e.g. 20Hz - 200Hz). Analysis of this data traction procedures, and a wide variety of (supervised) clas- in ubicomp is typically follows a pipeline-based approach sification methods [Bulling et al., 2014]. In many cases, [Bulling et al., 2014]. The first step is to segment the time- 1533 Figure 1: Models used in this work. From left to right: (a) LSTM network hidden layers containing LSTM cells and a final softmax layer at the top. (b) bi-directional LSTM network with two parallel tracks in both future direction (green) and to the past (red). (c) Convolutional networks that contain layers of convolutions and max-pooling, followed by fully-connected layers and a softmax group. (d) Fully connected feed-forward network with hidden (ReLU) layers. series data into contiguous segments (or frames), either driven been explored in two settings. First, [Neverova et al., 2015] by some signal characteristics such as signal energy [Plotz¨ investigated a variety of recurrent approaches to identify in- et al., 2012], or through a sliding-window segmentation ap- dividuals based on movement data recorded from their mo- proach. From each of the frames a set of features is ex- bile phone in a large-scale dataset. Secondly, [Ordo´nez˜ and tracted, which most commonly include statistical features or Roggen, 2016] compared the performance of a recurrent ap- stem from the frequency domain. proach to CNNs on two HAR datasets, representing the cur- The first deep learning approach explored to substitute for rent state-of-the-art performance on Opportunity, one of the this manual feature selection corresponds to deep belief net- datasets utilised in this work. In both cases, the recurrent net- works as auto-encoders, trained generatively with Restricted work was paired with a convolutional network that encoded Boltzmann Machines (RBMs) [Plotz¨ et al., 2011]. Results the movement data, effectively employing the recurrent net- were mixed, as in most cases the deep model was outper- work to only model temporal dependencies on a more ab- formed by a combination of principal component analysis stract level. Recurrent networks have so far not been applied and a statistical feature extraction process. Subsequently to model movement data at the lowest possible level, which is a variety of projects have explored the use of pre-trained, the sequence of individual samples recorded by the sensor(s). fully-connected networks for automatic assessment in Parkin- son’s Disease [Hammerla et al., 2015], as emission model 3 Comparing deep learning for HAR in Hidden Markov Models (HMMs) [Zhang et al., 2015; While there has been some exploration of deep models for Alsheikh et al., 2015], and to model audio data for ubicomp a variety of application scenarios in HAR there is still a applications [Lane et al., 2015]. lack of a systematic exploration of deep learning capabili- The most popular approach so far in ubicomp relies on con- ties. Authors report to explore the parameter space in prelim- volutional networks. Their performance for a variety of ac- inary experiments, but usually omit the details. The overall tivity recognition tasks was explored by a number of authors process remains unclear and difficult to replicate. Instead, [Zeng et al., 2014; Ronao and Cho, 2015; Yang et al., 2015; single instantiations of e.g. CNNs are presented that show Ronaoo and Cho, 2015]. Furthermore, convolutional net- good performance in an application scenario. Solely report- works have been applied to specific problem domains, such as ing peak performance figures does, however, not reflect the the detection of stereotypical movements in Autism [Rad et overall suitability of a method for HAR in ubicomp, as it re- al., 2015], where they significantly improved upon the state- mains unclear how much effort went into tuning the proposed of-the-art. approach, and how much effort went into tuning other ap- Individual frames of movement data in ubicomp are usu- proaches it was compared to. How likely is a practitioner to ally treated as statistically independent, and applications of find a parameter configuration that works similarly well for sequential modelling techniques like HMMs are rare [Bulling their application? How representative is the reported perfor- et al., 2014]. However, approaches that are able to exploit mance across the models compared during parameter explo- the temporal dependencies in time-series data appear as the ration? Which parameters have the largest impact on perfor- natural choice for modelling human movement captured with mance? These questions are important for the practitioner, sensor data. Deep recurrent networks, most notably those that but so far remain unanswered in related work. rely on Long Short-Term Memory cells (LSTMs) [Hochreiter In this paper we provide the first unbiased comparison of and Schmidhuber,
Load more

Recommended publications
  • Channel Selective Activity Recognition with Wifi: a Deep Learning Approach Exploring Wideband Information
    1 Channel Selective Activity Recognition with WiFi: A Deep Learning Approach Exploring Wideband Information Fangxin Wang, Student Member, IEEE, Wei Gong, Member, IEEE, Jiangchuan Liu, Fellow, IEEE and Kui Wu, Senior Member, IEEE Abstract—WiFi-based human activity recognition explores the correlations between body movement and the reflected WiFi signals to classify different activities. State-of-the-art solutions mostly work on a single WiFi channel and hence are quite sensitive to the quality of a particular channel. Co-channel interference in an indoor environment can seriously undermine the recognition accuracy. In this paper, we for the first time explore wideband WiFi information with advanced deep learning toward more accurate and robust activity recognition. We present a practical Channel Selective Activity Recognition system (CSAR) with Commercial Off-The-Shelf (COTS) WiFi devices. The key innovation is to actively select available WiFi channels with good quality and seamlessly hop among adjacent channels to form an extended channel. The wider bandwidth with more subcarriers offers stable information with a higher resolution for feature extraction. Conventional classification tools, e.g., hidden Markov model and k-nearest neighbors, however, are not only sensitive to feature distortion but also not smart enough to explore the time-scale correlations from the extracted spectrogram. We accordingly explore advanced deep learning tools for this application context. We demonstrate an integration of channel selection and long short term memory network (LSTM), which seamlessly combine the richer time and frequency features for activity recognition. We have implemented a CSAR prototype using Intel 5300 WiFi cards. Our real-world experiments show that CSAR achieves a stable recognition accuracy around 95% even in crowded wireless environments (compared to 80% with state-of-the-art solutions that highly depend on the quality of the working channel).
    [Show full text]
  • Human Activity Analysis and Recognition from Smartphones Using Machine Learning Techniques
    Human Activity Analysis and Recognition from Smartphones using Machine Learning Techniques Jakaria Rabbi, Md. Tahmid Hasan Fuad, Md. Abdul Awal Department of Computer Science and Engineering Khulna University of Engineering & Technology Khulna-9203, Bangladesh jakaria [email protected], [email protected], [email protected] Abstract—Human Activity Recognition (HAR) is considered a HAR is the problem of classifying day-to-day human ac- valuable research topic in the last few decades. Different types tivity using data collected from smartphone sensors. Data are of machine learning models are used for this purpose, and this continuously generated from the accelerometer and gyroscope, is a part of analyzing human behavior through machines. It is not a trivial task to analyze the data from wearable sensors and these data are instrumental in predicting our activities for complex and high dimensions. Nowadays, researchers mostly such as walking or standing. There are lots of datasets and use smartphones or smart home sensors to capture these data. ongoing research on this topic. In [8], the authors discuss In our paper, we analyze these data using machine learning wearable sensor data and related works of predictions with models to recognize human activities, which are now widely machine learning techniques. Wearable devices can predict an used for many purposes such as physical and mental health monitoring. We apply different machine learning models and extensive range of activities using data from various sensors. compare performances. We use Logistic Regression (LR) as the Deep Learning models are also being used to predict various benchmark model for its simplicity and excellent performance on human activities [9].
    [Show full text]
  • Human Activity Recognition Using Computer Vision Based Approach – Various Techniques
    International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056 Volume: 07 Issue: 06 | June 2020 www.irjet.net p-ISSN: 2395-0072 Human Activity Recognition using Computer Vision based Approach – Various Techniques Prarthana T V1, Dr. B G Prasad2 1Assistant Professor, Dept of Computer Science and Engineering, B. N. M Institute of Technology, Bengaluru 2Professor, Dept of Computer Science and Engineering, B. M. S. College of Engineering, Bengaluru ---------------------------------------------------------------------***---------------------------------------------------------------------- Abstract - Computer vision is an important area in the field known as Usual Activities. Any activity which is different of computer science which aids the machines in becoming from the defined set of activities is called as Unusual Activity. smart and intelligent by perceiving and analyzing digital These unusual activities occur because of mental and physical images and videos. A significant application of this is Activity discomfort. Unusual activity or anomaly detection is the recognition – which automatically classifies the actions being process of identifying and detecting the activities which are performed by an agent. Human activity recognition aims to different from actual or well-defined set of activities and apprehend the actions of an individual using a series of attract human attention. observations considering various challenging environmental factors. This work is a study of various techniques involved, Applications of human activity recognition are many. A few challenges identified and applications of this research area. are enumerated and explained below: Key Words: Human Activity Recognition, HAR, Deep Behavioral Bio-metrics - Bio-metrics involves uniquely learning, Computer vision identifying a person through various features related to the person itself. Behavior bio-metrics is based on the 1.
    [Show full text]
  • Vision-Based Human Tracking and Activity Recognition Robert Bodor Bennett Jackson Nikolaos Papanikolopoulos AIRVL, Dept
    Vision-Based Human Tracking and Activity Recognition Robert Bodor Bennett Jackson Nikolaos Papanikolopoulos AIRVL, Dept. of Computer Science and Engineering, University of Minnesota. to humans, as it requires careful concentration over long Abstract-- The protection of critical transportation assets periods of time. Therefore, there is clear motivation to and infrastructure is an important topic these days. develop automated intelligent vision-based monitoring Transportation assets such as bridges, overpasses, dams systems that can aid a human user in the process of risk and tunnels are vulnerable to attacks. In addition, facilities detection and analysis. such as chemical storage, office complexes and laboratories can become targets. Many of these facilities A great deal of work has been done in this area. Solutions exist in areas of high pedestrian traffic, making them have been attempted using a wide variety of methods (e.g., accessible to attack, while making the monitoring of the optical flow, Kalman filtering, hidden Markov models, etc.) facilities difficult. In this research, we developed and modalities (e.g., single camera, stereo, infra-red, etc.). components of an automated, “smart video” system to track In addition, there has been work in multiple aspects of the pedestrians and detect situations where people may be in issue, including single pedestrian tracking, group tracking, peril, as well as suspicious motion or activities at or near and detecting dropped objects. critical transportation assets. The software tracks individual pedestrians as they pass through the field of For surveillance applications, tracking is the fundamental vision of the camera, and uses vision algorithms to classify component. The pedestrian must first be tracked before the motion and activities of each pedestrian.
    [Show full text]
  • Fine-Grained Activity Recognition by Aggregating Abstract Object Usage
    Fine-Grained Activity Recognition by Aggregating Abstract Object Usage Donald J. Patterson, Dieter Fox, Henry Kautz Matthai Philipose University of Washington Intel Research Seattle Department of Computer Science and Engineering Seattle, Washington, USA Seattle, Washington, USA [email protected] {djp3,fox,kautz}@cs.washington.edu Abstract in directly sensed experience. More recent work on plan- based behavior recognition [4] uses more robust probabilis- In this paper we present results related to achieving fine- tic inference algorithms, but still does not directly connect grained activity recognition for context-aware computing to sensor data. applications. We examine the advantages and challenges In the computer vision community there is much work of reasoning with globally unique object instances detected on behavior recognition using probabilistic models, but it by an RFID glove. We present a sequence of increasingly usually focuses on recognizing simple low-level behaviors powerful probabilistic graphical models for activity recog- in controlled environments [9]. Recognizing complex, high- nition. We show the advantages of adding additional com- level activities using machine vision has only been achieved plexity and conclude with a model that can reason tractably by carefully engineering domain-specific solutions, such as about aggregated object instances and gracefully general- for hand-washing [13] or operating a home medical appli- izes from object instances to their classes by using abstrac- ance [17]. tion smoothing. We apply these models to data collected There has been much recent work in the wearable com- from a morning household routine. puting community, which, like this paper, quantifies the value of various sensing modalities and inference tech- niques for interpreting context.
    [Show full text]
  • Joint Recognition of Multiple Concurrent Activities Using Factorial Conditional Random Fields
    Joint Recognition of Multiple Concurrent Activities using Factorial Conditional Random Fields Tsu-yu Wu and Chia-chun Lian and Jane Yung-jen Hsu Department of Computer Science and Information Engineering National Taiwan University [email protected] Abstract the home for varying lengths of time. Sensor data are col- lected as the occupants interact with digital information in Recognizing patterns of human activities is an important this naturalistic living environment. As a result, the House n enabling technology for building intelligent home environ- ments. Existing approaches to activity recognition often fo- data set (Intille et al. 2006a) is potentially a good benchmark cus on mutually exclusive activities only. In reality, peo- for research on activity recognition. In this paper, we use the ple routinely carry out multiple concurrent activities. It is data set as the material to design our experiment. therefore necessary to model the co-temporal relationships Our analysis of the House n data set reveals a significant among activities. In this paper, we propose using Factorial observation. That is, in a real-home environment, people of- Conditional Random Fields (FCRFs) for recognition of mul- ten do multiple activities concurrently. For example, the par- tiple concurrent activities. We designed experiments to com- ticipant would throw the clothes into the washing machine pare our FCRFs model with Linear Chain Condition Random and then went to the kitchen doing some meal preparation. Fields (LCRFs) in learning and performing inference with the As another example, the participant had the habit of using MIT House n data set, which contains annotated data col- lected from multiple sensors in a real living environment.
    [Show full text]
  • Real-Time Physical Activity Recognition on Smart Mobile Devices Using Convolutional Neural Networks
    applied sciences Article Real-Time Physical Activity Recognition on Smart Mobile Devices Using Convolutional Neural Networks Konstantinos Peppas * , Apostolos C. Tsolakis , Stelios Krinidis and Dimitrios Tzovaras Centre for Research and Technology—Hellas, Information Technologies Institute, 6th km Charilaou-Thermi Rd, 57001 Thessaloniki, Greece; [email protected] (A.C.T.); [email protected] (S.K.); [email protected] (D.T.) * Correspondence: [email protected] Received: 20 October 2020; Accepted: 24 November 2020; Published: 27 November 2020 Abstract: Given the ubiquity of mobile devices, understanding the context of human activity with non-intrusive solutions is of great value. A novel deep neural network model is proposed, which combines feature extraction and convolutional layers, able to recognize human physical activity in real-time from tri-axial accelerometer data when run on a mobile device. It uses a two-layer convolutional neural network to extract local features, which are combined with 40 statistical features and are fed to a fully-connected layer. It improves the classification performance, while it takes up 5–8 times less storage space and outputs more than double the throughput of the current state-of-the-art user-independent implementation on the Wireless Sensor Data Mining (WISDM) dataset. It achieves 94.18% classification accuracy on a 10-fold user-independent cross-validation of the WISDM dataset. The model is further tested on the Actitracker dataset, achieving 79.12% accuracy, while the size and throughput of the model are evaluated on a mobile device. Keywords: activity recognition; convolutional neural networks; deep learning; human activity recognition; mobile inference; time series classification; feature extraction; accelerometer data 1.
    [Show full text]
  • LSTM Networks for Mobile Human Activity Recognition
    International Conference on Artificial Intelligence: Technologies and Applications (ICAITA 2016) LSTM Networks for Mobile Human Activity Recognition Yuwen Chen*, Kunhua Zhong, Ju Zhang, Qilong Sun and Xueliang Zhao Chongqing institute of green and intelligent technology Chinese Academy of Sciences *Corresponding author Abstract—A lot of real-life mobile sensing applications are features, while discrete cosine transform (DCT) have also been becoming available. These applications use mobile sensors applied with promising results [12], as well as auto-regressive embedded in smart phones to recognize human activities in order model coefficients [13]. Recently, time-delay embedding [14] to get a better understanding of human behavior. In this paper, have been applied for activity recognition. It adopts nonlinear we propose a LSTM-based feature extraction approach to time series analysis to extract features from time series and recognize human activities using tri-axial accelerometers data. shows a significant improvement on periodic activities The experimental results on the (WISDM) Lab public datasets recognition .However, the features from time-delay indicate that our LSTM-based approach is practical and achieves embedding are less suitable for non-periodic activities. The 92.1% accuracy. authors in [15] firstly introduce feature learning methods to the Keywords-Activity recognition, Deep learning, Long short area of activity recognition, they used Deep Belief Networks memory network (DBN) and PCA to learn features for activity recognition in ubiquitous computing. The authors in [16] following the work of [15] applied shift-invariant sparse coding technique. The I. INTRODUCTION authors in [17] also used sparse coding to learn features. Although human activity recognition (HAR) has been The features used in most of researches on HAR are studied extensively in the past decade, HAR on smartphones is selected by hand.
    [Show full text]
  • Convolutional Neural Networks for Human Activity Recognition Using Mobile Sensors
    WK,QWHUQDWLRQDO&RQIHUHQFHRQ0RELOH&RPSXWLQJ$SSOLFDWLRQVDQG6HUYLFHV 0REL&$6( Convolutional Neural Networks for Human Activity Recognition using Mobile Sensors Ming Zeng, Le T. Nguyen, Bo Yu, Ole J. Mengshoel, Jiang Zhu, Pang Wu, Joy Zhang Department of Electrical and Computer Engineering Carnegie Mellon University Moffett Field, CA, USA Email: {zeng, le.nguyen, bo.yu, ole.mengshoel, jiang.zhu, pang.wu, joy.zhang}@sv.cmu.edu Abstract—A variety of real-life mobile sensing applications are requires domain knowledge [23]. This problem is not unique to becoming available, especially in the life-logging, fitness tracking activity recognition. It has been well-studied in other research and health monitoring domains. These applications use mobile areas such as image recognition [22], where different types sensors embedded in smart phones to recognize human activities of features need to be extracted when trying to recognize a in order to get a better understanding of human behavior. While handwriting as opposed to recognizing faces. In recent years, progress has been made, human activity recognition remains due to advances of the processing capabilities, a large amount a challenging task. This is partly due to the broad range of human activities as well as the rich variation in how a given of Deep Learning (DL) techniques have been developed and activity can be performed. Using features that clearly separate successful applied in recognition tasks [2], [28]. These tech- between activities is crucial. In this paper, we propose an niques allow an automatic extraction of features without any approach to automatically extract discriminative features for domain knowledge. activity recognition. Specifically, we develop a method based on Convolutional Neural Networks (CNN), which can capture In this work, we propose an approach based on Convolu- local dependency and scale invariance of a signal as it has been tional Neural Networks (CNN) [2] to recognize activities in shown in speech recognition and image recognition domains.
    [Show full text]
  • Activity Recognition Using Cell Phone Accelerometers
    Activity Recognition using Cell Phone Accelerometers Jennifer R. Kwapisz, Gary M. Weiss, Samuel A. Moore Department of Computer and Information Science Fordham University 441 East Fordham Road Bronx, NY 10458 {kwapisz, gweiss, asammoore}@cis.fordham.edu ABSTRACT Mobile devices are becoming increasingly sophisticated and the 1. INTRODUCTION Mobile devices, such as cellular phones and music players, have latest generation of smart cell phones now incorporates many recently begun to incorporate diverse and powerful sensors. These diverse and powerful sensors. These sensors include GPS sensors, sensors include GPS sensors, audio sensors (i.e., microphones), vision sensors (i.e., cameras), audio sensors (i.e., microphones), image sensors (i.e., cameras), light sensors, temperature sensors, light sensors, temperature sensors, direction sensors (i.e., mag- direction sensors (i.e., compasses) and acceleration sensors (i.e., netic compasses), and acceleration sensors (i.e., accelerometers). accelerometers). Because of the small size of these “smart” mo- The availability of these sensors in mass-marketed communica- bile devices, their substantial computing power, their ability to tion devices creates exciting new opportunities for data mining send and receive data, and their nearly ubiquitous use in our soci- and data mining applications. In this paper we describe and evalu- ety, these devices open up exciting new areas for data mining ate a system that uses phone-based accelerometers to perform research and data mining applications. The goal of our WISDM activity recognition, a task which involves identifying the physi- (Wireless Sensor Data Mining) project [19] is to explore the re- cal activity a user is performing. To implement our system we search issues related to mining sensor data from these powerful collected labeled accelerometer data from twenty-nine users as mobile devices and to build useful applications.
    [Show full text]
  • Deep Learning Based Human Activity Recognition Using Spatio-Temporal Image Formation of Skeleton Joints
    applied sciences Article Deep Learning Based Human Activity Recognition Using Spatio-Temporal Image Formation of Skeleton Joints Nusrat Tasnim 1, Mohammad Khairul Islam 2 and Joong-Hwan Baek 1,* 1 School of Electronics and Information Engineering, Korea Aerospace University, Goyang 10540, Korea; [email protected] 2 Department of Computer Science and Engineering, University of Chittagong, Chittagong 4331, Bangladesh; [email protected] * Correspondence: [email protected]; Tel.: +82-2-300-0125 Abstract: Human activity recognition has become a significant research trend in the fields of computer vision, image processing, and human–machine or human–object interaction due to cost-effectiveness, time management, rehabilitation, and the pandemic of diseases. Over the past years, several methods published for human action recognition using RGB (red, green, and blue), depth, and skeleton datasets. Most of the methods introduced for action classification using skeleton datasets are con- strained in some perspectives including features representation, complexity, and performance. How- ever, there is still a challenging problem of providing an effective and efficient method for human action discrimination using a 3D skeleton dataset. There is a lot of room to map the 3D skeleton joint coordinates into spatio-temporal formats to reduce the complexity of the system, to provide a more accurate system to recognize human behaviors, and to improve the overall performance. In this paper, we suggest a spatio-temporal image formation (STIF) technique of 3D skeleton joints by capturing spatial information and temporal changes for action discrimination. We conduct transfer learning (pretrained models- MobileNetV2, DenseNet121, and ResNet18 trained with ImageNet Citation: Tasnim, N.; Islam, M.K.; dataset) to extract discriminative features and evaluate the proposed method with several fusion Baek, J.-H.
    [Show full text]
  • Human Activity Recognition Using Deep Learning Models on Smartphones and Smartwatches Sensor Data
    Human Activity Recognition using Deep Learning Models on Smartphones and Smartwatches Sensor Data Bolu Oluwalade1, Sunil Neela1, Judy Wawira2, Tobiloba Adejumo3 and Saptarshi Purkayastha1 1Department of BioHealth Informatics, Indiana University-Purdue University Indianapolis, U.S.A. 2Department of Radiology, Imaging Sciences, Emory University, U.S.A. 3Federal University of Agriculture, Abeokuta, Nigeria Keywords: Human Activities Recognition (HAR), WISDM Dataset, Convolutional LSTM (ConvLSTM). Abstract: In recent years, human activity recognition has garnered considerable attention both in industrial and academic research because of the wide deployment of sensors, such as accelerometers and gyroscopes, in products such as smartphones and smartwatches. Activity recognition is currently applied in various fields where valuable information about an individual’s functional ability and lifestyle is needed. In this study, we used the popular WISDM dataset for activity recognition. Using multivariate analysis of covariance (MANCOVA), we established a statistically significant difference (p < 0.05) between the data generated from the sensors embedded in smartphones and smartwatches. By doing this, we show that smartphones and smartwatches don’t capture data in the same way due to the location where they are worn. We deployed several neural network architectures to classify 15 different hand and non-hand oriented activities. These models include Long short-term memory (LSTM), Bi-directional Long short-term memory (BiLSTM), Convolutional Neural Network (CNN), and Convolutional LSTM (ConvLSTM). The developed models performed best with watch accelerometer data. Also, we saw that the classification precision obtained with the convolutional input classifiers (CNN and ConvLSTM) was higher than the end-to-end LSTM classifier in 12 of the 15 activities.
    [Show full text]