ECG Based Recognition Using Second Order Statistics

This paper investigates the applicability of electrocardiogram (ECG) signals for human recognition. Current approaches apply feature extraction on a fiducial points basis. In this paper we demonstrate an autocorrelation based feature extraction approach, in conjunction with the discrete cosine transform or linear discriminant analysis. As an optimization, we introduce a template matching technique that substantially improves the classification performance while also acting as an intruder detector. The experimental results show considerably high recognition rates, rendering identification applications based on ECG very promising.