Biometric personal identification system using the ECG signal

The electrocardiogram has unique cardiac features to each individual, which motivated us to use it as a biometric, hence, its robust nature against falsification makes it rather reliable for security systems, as it offers ultimate security in all situations. This paper presents a new approach applying this ECG particularity. A robust ECG Biometrics based on the features extraction with fiducial detection in the time domain is proposed. After preprocessing, ten morphological descriptors are extracted from each heartbeat and which were divided into homogenous groups (amplitude, surface, interval and slope). Later, sixty Hermite Polynomials Expansion (HPE) coefficients are extracted from the ECG signal. Finally, classification is based on the Hidden Markov Model (HMM) with the (HTK) recognition toolkit using a Bakis model with one Gaussian. A particular strategy is adopted for personal identification: the groups of morphological parameters were used separately in the classification then were made associations between these groups them in one input vector. On the other hand, the Hermite Polynomials Expansion coefficients were classified apart. In order to improve performance, a combination between 10 morphological parameters and 60 HPE coefficients was performed in a one system. Our algorithm is tested on 18 healthy signals of the MIT BIH database. Each recording lasts about 30 minutes, 20 min of the data were used for training and the last 10 min were used for testing. The analysis of different groups separately showed that the best recognition performance is 96.7% for the Hermite Polynomials Expansion coefficients and the results of experiments showed that the proposed hybrid approach has led to an overall maximum of 99%.