ECG to Individual Identification

This paper presents an individual identification system using single lead electrocardiogram (ECG). The proposed techniques for P and T wave delineation are based on time derivative and adaptive thresholding. The performance of proposed delineators is evaluated on manually annotated Physionet QT database. The accuracy of delineators are quantified on mean error and standard deviation of differences between manually annotations and automated results. Especially, lower values of error in standard deviation for onset and offset of P wave fiducials are obtained as 8.1 and 6.29 while for T wave fiducials are 9.4 and 11.2 (where units are in ms). It shows the performance of P and T wave delineators is optimum and also stable in comparison to other published results. Found fiducials are processed for the extraction of heartbeat features. From each heartbeat, 19 stable features related to interval, amplitude and angle are computed. The feasibility of ECG as a new biometric is tested on proposed identification system designed on template matching and adaptive thresholding. The accuracy of identification system is achieved to 99% on the datasize of 125 recordings prepared from 25 individual ECG of Physionet.