Soft Biometrics: an Asset for Personal Recognition

Soft Biometric: An Asset for Personal Recognition

Chander Kant [1], Rajender Nath [2]

[1] Lecturer, [2] Reader

Department of Computer Science and Applications K.U., Kurukshetra, Haryana, INDIA

[ckverma, rnath_2k3]@ rediffmail.com

Abstract

Biometric systems automatically recognize individuals based on their physiological and behavioral characteristics. Hence, the fundamental requirement of any biometric recognition system is a human trait having several desirable properties like universality, distinctiveness, permanence, acceptability. However, a human characteristic that possesses all these properties has not yet been identified. As a result, none of the existing biometric systems provide perfect recognition and there is a scope for improving the performance of these systems. Although characteristics like gender, age, height, eye color and weight are not unique and reliable, they provide some information about the user. We refer to these characteristics as “soft” biometric traits and argue that these traits can be integrated with the identity information provided by the primary biometric identifiers like fingerprint, face, iris, signature etc. We propose the utilization of “soft” biometric traits like gender, height, weight, age, and ethnicity to complement the identity information provided by the primary biometric identifiers. Although soft biometric characteristics lack the distinctiveness and permanence to identify an individual uniquely and reliably yet they provide some evidence about the user identity that could be beneficial. This paper presents a framework for integrating the ancillary information (soft biometrics) with the primary biometric system. Algorithm based of existing data and facts show that the recognition performance of an identification system can be improved significantly by using additional user information like gender, ethnicity, and height.

Keywords: primary biometric, soft biometric, universality, distinctiveness, permanence, acceptability.

1. Introduction

There are a number of methods to verify identity adopted by society or automated systems. These are summarized in Table1. We define soft biometric traits as characteristics that provide some information about the individual, but lack the distinctiveness and permanence to sufficiently differentiate any two individuals. The soft biometric traits can either be continuous (e.g., height and weight) or discrete (e.g., gender, eye color, ethnicity, etc.). In this paper, we describe a framework for integrating the information provided by the soft biometric with the primary biometric system. Biometric

systems recognize users based on their physiological and behavioral characteristics [1]. Unimodal biometric systems make use of a single biometric trait for user recognition.

Methods / Examples / Problems
What you Know / Password, PIN, ID / Forgotten, Shared, easy to guess
What you have / Key, Cards, etc / Lost or Stolen, Can be duplicated
What you are / Fingerprint, Face, Iris… / Non-repudiable authentication

Table 1: Identification Technologies

It is difficult to achieve very high recognition rates using unimodal systems due to problems like noisy sensor data and non-universality or lack of distinctiveness of the chosen biometric trait. Multimodal biometric systems address some of these problems by combining evidence obtained from multiple sources [2]. A multimodal biometric system that utilizes a number of different biometric identifiers like face, fingerprint, hand-geometry, and iris can be more robust to noise and minimize the problem of non-universality and lack of distinctiveness. However, a multimodal system will require a longer verification time thereby causing inconvenience to the users. A multimodal biometric system based on different biometric identifiers can be expected to be more robust to noise, address the problem of non-universality, improve the matching accuracy, and provide reasonable protection against spoof attacks. However, such a system will have two limitations. Firstly, the overall cost involved in building the multimodal system can be high due to the need for multiple high quality sensors and increased storage and computational requirements. Secondly, the system will require a longer verification time thereby causing inconvenience to the users. Due to these limitations, the number of identifiers in a multimodal biometric system is usually restricted to two or three.

A possible solution to the problem of designing a reliable and user-friendly biometric system is to use additional information about the user like height, weight, age, gender, ethnicity, and eye color to improve the performance of the primary biometric system. Most practical biometric systems collect such additional information about the users during enrollment. This information is stored either in the database or in the smart cards possessed by the user. However, this information is not currently utilized during the automatic identification/verification phase. This motivates us to utilize every available information about the user to improve the performance of a biometric recognition system. Further, biometric systems used in access control applications generally have a human supervisor who oversees the operations of the system. When a genuine user is falsely rejected by the system, the human operator steps in to verify the identity of this user. This manual verification is usually done by comparing the facial appearance of the user with the facial image appearing on the user’s identification card and by verifying other information on the ID card like age, gender, height, and other visible identification marks. If the soft biometric characteristics can be automatically extracted and used during the decision making process, the overall performance of the system will improve and the need of manual involvement will be reduced.

2. Related Work

Integrating Soft Biometrics with Primary Biometrics to improve FAR and FRR.

The first personal identification system developed by Alphonse Bertillon [3] for identification of criminals was based on three sets of features: (i) Anthropometric measurements like height and length of the arm, (ii) Morphological description of the appearance and body shape like eye color and anomalies of the fingers, and (iii) Peculiar marks observed on the body like moles and scars. Although the Bertillon system was useful in tracking criminals, it had an unacceptably high error rate because the features used are indistinctive and non-permanent. Heckathorn et al. [4] have shown that a combination of personal attributes like age, gender, eye color, height, and other visible identification marks can be used to identify an individual only with a limited accuracy. Hence, a system that is completely based on soft biometric traits cannot meet the accuracy requirements of real world applications. However, soft biometric traits can be used to improve the performance of traditional biometric systems.

Wayman [5] proposed the use of soft biometric traits like gender and age, for filtering a large biometric database (Figure1). Filtering refers to limiting the number of entries in a database to be searched, based on characteristics of the interacting user. For example, if the user can somehow be identified as a middle-aged male, the search can be restricted only to the subjects with this profile enrolled in the database. This greatly improves the speed or the search efficiency of the biometric system. Filtering and system parameters tuning require an accurate classification of a user into a particular class (e.g., male or female, blue or brown eyes, Caucasian or Asian or African). This requires a pre-identification module that can accurately perform this classification. The probability of user ώi, given by primary biometric feature vector x and the soft biometric feature vector y i.e., P(ώi | x; y) can be calculated using the Bayes’ rule.

P(ώi|x; y) = p(y|ώi) P(ώi|x)

------

for i=1 to n p(y|ώi) P(ώi|x)

where P(ώi | x) is the probability that the test user is ώi given the feature vector x. If the output of the primary biometric system is a matching score, it is converted into posteriori probability using an appropriate transformation. For the secondary biometric system, we can consider P(ώi | x) as the prior probability of the test user being user ώi.

The logarithm of P(ώi|x; y) in equation (3) can be expressed as

This formulation has two main drawbacks. The first problem is that all the m soft biometric variables have been weighed equally. For example, the gender of a person may give more information about a person than height. Therefore, we must introduce a weighting scheme for the soft biometric traits.

Figure 2. Improvement in authentication performance after utilization of soft biometric traits.

Another problem is that any impostor can easily spoof the system because the soft characteristics have an equal say in the decision as the primary biometric trait. It is relatively easy to modify/hide one’s soft biometric attributes by applying cosmetics and wearing other accessories (like mask, shoes with high heels, etc.). We can overcome to the above mentioned problem with some way [9] and can get the better result as shown in figure [2].

3. Proposed work

Integrating Soft Biometrics with Primary Biometrics to reduce process time

In our framework, the biometric recognition system is divided into two subsystems. One subsystem is called the primary biometric system and it is based on traditional biometric identifiers like fingerprint, face and hand-geometry [7]. The second subsystem, referred to as the secondary biometric system, is based on soft biometric traits like age, gender, and height. Figure [3] shows the architecture of a personal identification system that makes use of both fingerprint and soft biometric measurements. The combination of such system is also discussed earlier as discussed in previous section of this paper but here we have put soft biometrics as first stage and output of first stage is fed into second stage that is based on primary biometric system. This type of combination will reduce the processing time system and will improve the efficiency of system. Processing at Stage-II will be performed only when the user will pass the stage-I else there will be rejection automatically even without computing the stage-II. Stage-I automatically computes the soft traits like age, gender, height of user.

(Figure3. Architecture of Personal Identification using Primary Biometrics and Soft Biometric)

Experimental results show significant improvement in recognition performance due to the utilization of soft biometric information. We used fingerprint as the primary biometric identifier and age, gender, and height as the soft biometric variables. We have chosen soft traits in such a way that they could be easily computed automatically.

Let us illustrate the improvement in Total Response Time with the help of an example. Suppose, there are 100000 templates to be compared and matched and out of these 100000 templates there are only 100 templates which are supposed to be matched against the input templates. Let us assume that time taken by Primary Biometric System = TF and time taken by Soft Biometric = TS (where TF>TS around 10 times, because soft traits are easily calculated at fast rate [6]. In conventional Fingerprint verification system, Total Processing Time TR = 100000*TF.

In Proposed system Tp = (100000-100)* TS + 100 *(TF + TS).

(Since 99900 templates are discarded at Level-I. Only 100 templates are need verified at both of levels).If ratio of TF: TS is 10: 1 then

TR = 100000*10=1000000 and

Tp = 99900 + 100(10+1) = 99900+1100=101000.

Ratio R= TR / Tp = 1000000/101000= 9.9 times that is apx 10 times.

i.e. total process time of proposed system increases 10 times compared to conventional system. Further this ratio becomes more when more templates are taken at input side. This paper presents an approach that improves the processing time of existing conventional fingerprint verification system. The approach is useful when fingerprint verifications are made at large scale level (off-line). The proposed scheme is not free from all drawbacks, as it needs extra storage space to store the templates with soft trait data like age, gender, height.

4. Extraction of Soft Biometric

In order to utilize soft biometrics, there must be a mechanism to automatically extract these features from the user during the recognition phase. As the user interacts with the primary biometric system, the system should be able to automatically measure the soft biometric characteristics like height, age and gender without any interaction with the user. This can be achieved using a special system of sensors. For example, a bundle of infra-red beams could be used to measure the height. A camera could be used for obtaining the facial image of the user, from which information like age, gender, and ethnicity could be derived [8]. These observed soft biometrics information could then be used to supplement the identity information provided by the user’s primary biometric identifier. Extensive studies have been made to identify the gender, ethnicity, and pose of the users from their facial images. Gutta et al.[10] proposed a mixture of experts consisting of ensembles of radial basis functions for the classification of gender, ethnic origin, and pose of human faces. Their gender classifier classified users as either male or female with an average accuracy rate of 96%. Age determination is a more difficult problem due to the very limited physiological or behavioral changes in the human body as the person grows from one age group to another. There are currently no reliable biometric indicators for age determination.

5. Summary and Future Prospectus

We have demonstrated that the utilization of ancillary user information like gender, height, and ethnicity can improve the performance of the traditional biometric systems. Although the soft biometric characteristics are not as permanent and reliable as the traditional biometric identifiers like fingerprint, they provide some information about the identity of the user that leads to higher accuracy in establishing the user identity. We have also shown that soft biometric characteristics would help only if they are complementary to the primary biometric traits. However, an optimum weighting scheme based the discriminative abilities of the primary and the soft biometric traits is needed to achieve an improvement in recognition performance. Our future work in this direction will involve the improvement of proposed system free from the problems discusses in section 3. Also more accurate mechanism must be developed for automatic extraction of soft biometric traits.

6. References

[1] Jain, A.K., Bolle, R., Pankanti, S., eds.: Biometrics: Personal Identification in Networked Security. Kluwer Academic Publishers (1999).

[2] Hong, L., Jain, A.K., Pankanti, S.: Can Multibiometrics Improve Performance? In: Proceedings of IEEE Workshop on Automatic Identification Advanced Technologies, New Jersey, U.S.A. (1999) 59–64