Individual recognition by gaussian ECG features

Authentication security is a wide and global problem. Target user authentication systems based on biometric features are less vulnerable than traditional methods. In particular, ECG signal has the advantage of being concealed, therefore it is more difficult to replace or stole. In this work, we propose an approach based on ECG Gaussian features extracted by modeling each cardiac cycle wave with Gaussian kernels. Model parameters are the elements of feature vector. The authentication step is supported by One Class Support Vector Machine (OCSVM), specifically designed and optimized for the problem under consideration. The authentication has been made more robust with a Majority Vote system based on the consideration of several cardiac cycles. The performance of the proposed system is evaluated on the subjects of MIT/BIH Physionet Database. The results show that the proposed method performs better than other approaches reported in the literature and achieves an overall performance in terms of F-score of 98.44%.