Biometric identification of individuals based on the ECG. Which conditions?

Biometric systems have for objective to perform identification, or identity verification, of individuals. Human ECG has been recently proposed as an additional tool for biometric applications. However, most of the existing studies are set in supine rest and only consider the QRS morphology after, most of the time, feature extraction from the ECG. This paper is focused on identification based on pattern recognition by comparing ECG shapes. Experiments were conducted on a database containing 11 healthy subjects, recorded in three different experimental conditions (supine rest, standing, exercise) and repeated up to four times, over 16 months. We calculate the correlation coefficient between a shape coming from an unknown individual and all the shapes of the database. We also evaluate the influence of the recording condition, the shape length and the number of leads. Best results (100% of good identification in the first rank) are obtained using the whole enrolment database, 12 leads and a shape length of 500 ms.