Title: Human identification system based on ECG features

Authors: Dhouha Rezgui; Zied Lachiri

Addresses: Department of Electrical Engineering, LR-SITI Laboratory, National School of Engineers of Tunis, Le Belvédère, Tunis 1002, Tunisia ' Department of Electrical Engineering, LR-SITI Laboratory, National School of Engineers of Tunis, Le Belvédère, Tunis 1002, Tunisia

Abstract: Currently biometrics is widely used for security needs. In particular, the ECG that provides optimum security since it is more universal and difficult to be forged. This research presents a new approach for personal identification from electrocardiogram signals. After pre-processing, fiducial points were detected for each heartbeat. Then three types of features, namely temporal attributes, amplitude attributes and morphological descriptors, are extracted. Hidden Markov model was used for analysis of parameters and personal recognition. A combination between 21 features and 10 morphological parameters was performed in a one system in order to bring more significant improvement in terms of recognition. Results demonstrate that the proposed method is efficiently used to identify the normal and diseased subjects. In particular, the best identification rate of 99.10% is obtained for the subjects of MIT-BIH Normal Sinus Rhythm database whereas the subjects of MIT-BIH Arrhythmia database have led to a recognition accuracy of 98.39%.

Keywords: biomedical engineering; ECG signals; electrocardiograms; temporal features; amplitude features; morphological parameters; HMM; hidden Markov models; MIT-BIH database; human identification systems; biometrics; personal identification.

DOI: 10.1504/IJBET.2015.071412

International Journal of Biomedical Engineering and Technology, 2015 Vol.19 No.1, pp.92 - 103

Received: 17 Jan 2015
Accepted: 18 Mar 2015
Published online: 25 Aug 2015 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article