Personal recognition using geometric features in the phase space of electrocardiogram

Electrocardiogram (ECG) is one of the biomedical properties that have recently been studied for biometrics. We prop ose a new method for human recognition by phase-space reconstruction (PSR) of a single-lead ECG signal. To reconstruct a single-lead ECG signal into phase space, we used a time delay technique We extracted the geometric features through the trajectory from the phase space and analyzed it using our method. The ECG signals used in this study were measured in various situations such at rest, during exercise, while listening to music, and watching a video. We performed phase space reconstruction by applying some time-delay to the measured ECG signal and extracted 21 geometric al features to find the best identifiable time-delay value through Support vector machine learning. The results were performed on 1 3 subjects. The accuracy was 97.8% when the delay was 8ms. Based on this result personal authentication was conducted. The results show 97.7% accuracy, 1.5% FAR(False Acceptance Ratio) and 2.9% FRR(False Rejection Ratio).