Research On The Feature Extraction And Recognition Of Pulse Signals

With the development of Chinese medicine, pulse-diagnosis as a non-invasive mean and method of detection has gained much appreciation and concerns of domestic and foreign people , but since ancient times , Chinese medicine has been obtaining the pulse of information from fingers . Although the pulse diagnosis is of simple, non-invasive and painless , and can be easily accepted by patients , has also revealed some shortcomings in the long-term medical practice. With the development of sensor and computer processing technology, it has been a hot topic in domestic and overseas scholars to realise the objectification of pulse-diagnosis, study the detection of pulse, and to promote the modernization of pulse-diagnosis by modern technology and equipment, which is also the research purpose of this paper.This paper focuses on the basic concepts and fundamental theory of feature extraction and selection in detail, and discusses separability criterion. At the same time, multiresolution analysis to get the clear idea of the scalar coefficients, compared to analysis of experiment result by extracting the different characteristic parameter.Wavelet transform is a good analytical method both in the time and the frequency domains, That is, in the low-frequency part of the high frequency resolution and lower temporal resolution, high-frequency part of a high time resolution and lower frequency resolution, so as the "mathematical microscope" , especially applicable for non-stationary signal processing. In this paper we analyze pulse signals of 20 heroin addicts and 20 healthy persons using the multiresolution analysis of wavelet transform and Euclidean distance. Extracted through the scalar coefficients of the wavelet transform,By means of calculating the scalar coefficients of the wavelet transform to the central square of the distance category, we found the significant difference between the heroin addicts and the healthy persons, a primary criterion for measuring off the heroin addicts and the healthy persons was obtained. Based on this criterion, the 20 healthy persons were identified and 1 heroin addicts were misjudged.This paper also deduced the theorems and formulas of the Support Vector Machines(SVM)and gave especial research on algorithm of the the optimal facial classification, which is much helpful for the model recognition. At the basis of the feature extraction of pulse signals, this paper also uses the Support Vector network to identify the 40 pulse signals. (20 heroin druggers, 20 normal healthy people) ,to be given the optimal facial classification. And to achieve the desired results.