PPG-Based Biometric Recognition


Medical practitioners often monitor the functioning of human body by non-invasively collecting bioelectric signals such as electrocardiogram (ECG), electroencephalogram (EEG), and photoplethysmogram (PPG). These signals are used to examine the health of human beings and diagnose different ailments in them. However, in recent years, in addition to their obvious application in medical science, researchers have extensively used these bioelectric quantities for biometric recognition. Traditionally, the area of biometrics has involved physiological traits like fingerprint, iris, face etc. or behavioural qualities like gait, keystroke, speech, etc. for distinguishing individuals. The USP of biometric lies in the fact that being an integral part of a person, these traits cannot be stolen, shared, or forgotten. The uniqueness and permanence of these traits, especially the physiological ones, are well established and technology linked to them has also matured. Primarily, these tools have been employed in access control and security applications. The systems operate either in verification mode, where one-to-one matching takes place, or identification mode, which involves many-to-one comparison. However, with widespread digitisation, the scope of applications of biometrics has also widened to varied domains ranging from e-commerce to healthcare [1].

With the enhanced popularity and visibility of biometric systems, there has been a rapid upsurge in attempts by the fraudsters to breach these systems and intrude into them. The different points/stages at which biometric system can be intruded are depicted in Figure 13.1.

These can be divided into two main groups: (i) direct attacks involve use of synthetic templates such as pre-recorded speech, face images etc and are at sensor level (attack 1), and (ii) indirect attacks consist of all the remaining errors. Out of these, feature extractor and matcher are bypassed using a Trojan horse in stages 3 and 5, respectively. The stored template is replaced, added, or deleted in attack at stage 6. While weaknesses in the communication channel are exploited by attacks at stages 2, 4, 7, and 8. Indirect attack is basically an insider attack, as intruder should have a significant knowledge about the working of the system.

A solution to the problem of ‘direct attack’ has been provided by adopting a multimodal scheme, i.e. the use of multiple traits like fingerprint and face or fingerprint and speech in a single system. However, in many of these combinations, also, the liveliness component is still missing and an additional or special hardware needs to be attached to ensure that the samples have been obtained from alive individuals. Another remedy is to use a trait which has vitality property intrinsically embedded in it, and this led the researchers to test the efficacy of bioelectric signals like ECG, EEG, and PPG, etc. for biometric applications. These signals are naturally present in all human beings and also have an in-built trait for vitality check. In the last two decades, a considerable amount of work has been done in this upcoming area. Among them, the acquisition of PPG is most user-friendly and least intrusive. This makes it more suitable for integration with other biometric traits in a multimodal system. This chapter provides a review of PPG-based biometric recognition along witli the review of research works applying new age machine learning and deep learning techniques in this area [2].

The remaining chapter contains four more sections with Section 13.2 providing an overview of PPG signal and the associated terminology. The literature related to PPG-based human recognition has been presented in Section 13.3. A scheme for person recognition using PPG and the results obtained are discussed in Section 13.4. Finally, the chapter ends with Section 13.5 in which the conclusions are given.

Possible attack points in a typical biometric system

FIGURE 13.1 Possible attack points in a typical biometric system.

A typical PPG pulse and its characteristic points

FIGURE 13.2 A typical PPG pulse and its characteristic points.


Photoplethysmograph, made up of photo (light) plus two Greek words plethysmos (increasing) and graph (write), is an optical instrument through which the changes in the blood volume can be detected and measured. This non-invasive method also known as photoelectric plethysmography was introduced by Alrick Hertzman in late 1930s. The signal acquired by using this instrument is called photoplethysmogram (PPG). A PPG signal is captured using an optical sensor, emitting red and infrared light, placed on the fingertip, earlobe, toe-tip, etc., with the first two acquisition sites being more popular and convenient. PPG signal shows the change in the volume of blood at the site of acquisition (fingertip) as the heart pumps the blood to various extremities. As shown in Figure 13.2, PPG pulse comprises of two waves: one systolic and the other diastolic wave [3].

Predominantly, PPG has been employed for medical applications like measuring saturation of oxygen, blood pressure, cardiac output, etc. However, in the last one- and-half decade, researchers have explored the possibility of utilising this signal for human authentication and identification. A review of literature in this direction is presented in the next section.

< Prev   CONTENTS   Source   Next >