Human recognition using discrete cosine transform and discriminant analysis of ECG

The electrocardiogram (ECG) is a bio electrical signal generated from heart of an individual. It does not require vitality testing if used as biometrics. The ECG biometric systems are robust and have potential to prevent the fraudulent attacks. This paper explores a novel method of human recognition based on autocorrelation (AC) and linear discriminant analysis (LDA). AC is used to extract features of ECG wave without relying on the accuracy of dominant fiducial point extraction. These features are transformed to its discrete cosine coefficients to maximize similarities and differences among the subjects. Further, the performance of two dimensionality reduction techniques ie. principal component analysis (PCA) and LDA are evaluated on transformed feature vector. The experimental results show that the proposed method achieves the equal error rates (EER) of 0.90% and 0.06% while using LDA on MIT-BIH arrhythmia and QT database respectively. It outperforms the other results on these databases.