Individual Verification based on Difference Parametric Features Extracted from P and QRS waves of ECG Signal

Feature extraction with high discrimination ability is key to authentication systems. This paper illustrates information on Extracting DCT coefficients and parametric features of biometric electrocardiogram-P and QRS waves-for individual verification. The proposed solution is based on a reference features and producing difference characteristics ,and thus increasing the discriminate ability classifier and achieving beter accuracy. It was applied on prepared database include 28 healthy peaple. Classification of genuine and forgery patterns with using four methods of K-Nearest Neighbor, Least Square Error, Gaussian Mixture Model, and Fuzzy K-Nearest Neighbor are implemented. The test result indicate that using K-Nearest Neighbor classification with 20 superior parametric and DCT features has allowed to achieve equal error rate 1.13%±0.63 with an accuracy of 98.87%.

کلیدواژه ها:

biometric ، ECG ، feature selection ، authentication ، classification