Biometric Based on Parametric Features of Electrocardiogram Signal

Biometric Based on Parametric Features of Electrocardiogram Signal

Using biological signals to biometrics is into consideration today. Among the crucial advantages of using the biological signals we can refer to its high security and people validity cannot simply fake it, and also as long as the person is alive, itʹs accessible and usable. In this paper, the cardiac signal is used to confirm the identity. Signal processing involves the steps of signal filtering; segmentation and extraction of 20 parametric characteristics and selection of 5, 10 and 15 superior features. Classification of real and unreal patterns with the help of four methods of K-Nearest Neighbor, Least Square Error, Gaussian Mixture Model, and Fuzzy K-Nearest Neighbor are implemented. The test results indicate that using fuzzy K-nearest neighbor classification with 15 superior features and combination of artificial characteristics of the average, minimum and maximum of basic characteristics, has allowed to achieve equal error rate, %1.99 ± 0.38 with an accuracy of %98.1 ± 0.38.