ECG biometric analysis in different physiological recording conditions

Abstract

Biometric systems have for objective to perform identification or verification of identity of individuals. Human electrocardiogram (ECG) has been recently proposed as an additional tool for biometric applications. Then, a set of ECG-based biometric studies has occurred in the literature, but they are difficult to compare because they use various values of: the number of ECG leads, the length of the analysis window (only the QRS or more), the delays between recordings\(\ldots \) However, they analyse nearly always the ECG in rest conditions. Here, we propose to evaluate the possibility of performing ECG-based biometry in other conditions. For this purpose, a comparative study, on three experimental conditions (supine rest, standing and exercise), has been carried out. It is based on the computing of the correlation coefficient between pairs of shapes of windowed ECG. Both verification and identification tasks are tackled. The results show that there is no advantage in comparing shapes recorded in supine rest conditions, as it is classically done, which represents an obvious benefit in biometry. Performances are evaluated as a function of the shape length. Then, different tests are performed in order to investigate how should be constructed the enrolment database when a system is devoted to work in several conditions. Last part of the paper shows how performances depend on the time and on the number of ECG leads.