Automobile Driver Recognition under different physiological conditions using the electrocardiogram

This paper presents a person identification mechanism of automobile drivers under different physiological conditions. A total of 16 subjects were used in this study from the Stress Recognition in Automobile Driver database (DRIVEDB). Discrete Wavelet Transform was applied to reveal useful hidden information in the ECG signal which is not readily available in a time domain representation. Features are extracted based on coefficients produced due to the wavelet decomposition process. These features sets were then used in Radial Basis Function (RBF) for classification purposes. Our experimentation suggests that person identification is possible by obtaining identification accuracy of 95% as compared to 91% without wavelet analysis. This also indicates the robustness of ECG biometric implemented under different physiological conditions.