Based On The Wavelet Transform Of Ecg Feature Extraction And Classification Recognition Research

Biometrics refers to automatic personal identification on the basis of some physiological characteristics or some behavioral aspects. Because of its many outstanding advantages what other conventional verification methods lack, biometrics has gradually become a hot research topic in the world. Now, several biometrics that have been used commercially for human identity verification are facial geometry, fingerprints and voice analysis. Unfortunately, people often leave fingerprints on objects which are easy to be stolen; face shape can be picked up easily from a photo; voice can by imitated and so on. So researching the new biometrical identification methods is still one of the focuses in the field of human identity verification.Electrocardiogram (ECG) analysis is not only a very useful diagnostic tool for clinical proposes but also a good biometric for human identification. Firstly, it has universality that is to say every person Possesses it; Secondly, the ECG varies from Person to Person duing to the differences in position, size, anatomy of the heart, age, gender, relative body weight, chest configuration and various other factors. Therefore, it is unique. Thirdly, people’s ECG remains unchangeable during a longer time so it is stable. Recently people have begun to use ECG as a means of human identity verification and have got some results.Because the ECG signal as a fairly new identity verification technology is still in the initial stage which is need to be improved. In order to improve the veracity and the efficiency of identification algorithms, new work has been done as follows based on the basis of previous studies:(1) Take the MIT-BIH arrhythmia database as the feature extraction data to testing the method validity. Choose the50cases ECG signal come from different five personal to identification of ECG experiment data.(2) To consider the original signal containing the frequency interference and baseline drift, we used Butterworth filter and median filter to remove the interference in ECG signal.(3) We introduced the approach of using the wavelet transform in feature extraction. Adopting this method we extracted15time distance characteristics and6amplitude characteristics which would be used during the identity verification. The experiments proved wavelet transform had more advantages than traditional differential threshold methods during QRS locating in the calculation speed and location accuracy.(4) An introduction of BP network and deduction of BP algorithm were introduced. ECG identification was achieved based on the characteristic parameters by constructing and training the part of sample series.Experimental results showed that characteristics recognition of ECG is a operational biological recognition method and the identification rate of five people reached up to100%with the BP network methods. From the result we can see that the ECG signal identification method is a new kind of promising identification method.