Personal identification system using physiological signal

The heart electrical activity has highly distinctive characteristics for each individual, which motivated us to use the electrocardiogram signal (ECG) as a biometric characteristic. This paper presents a new approach for human recognition from their Electrocardiograms (ECG). The proposed approach is based on an extraction feature with fiducial detection in time domain. It consists of three steps. After the preprocessing step, we extracted ten morphological descriptors and sixty hermite Polynomials Expansion coefficients (HPEc) from each heartbeat which represented the second step. The classification step is done by using a binary SVM and the hidden Markov model (HMM). In this step, a particular strategy of persons' identification was adopted: Firstly, we used four groups of morphological descriptors separately. Secondly, we combined them in one input vector of our classification system. Moreover, we employed the hermite Polynomials Expansion coefficients apart. In order to improve performance, we associated the HPEc with all groups of morphological descriptors as input parameters of our classification system. We tested our algorithm on 18 healthy signals of the MIT_BIH database. The results of experiments showed that the proposed hybrid approach has achieved an identification rate equal to 98.97% using SVM and 99.02% using HMM.