Locating optimized feature points of human pulse based on support vector regression

Traditional Chinese Pulse Diagnosis (TCPD), one of the four diagnostic methods of Traditional Chinese Medicine (TCM), had been proved to be clinically valid in Chinese Medicine history. But traditional pulse diagnosis is subjective and deficient in quantitative criteria of diagnosis, which affects the reliability and repeatability of pulse diagnosis. Therefore, quantitative methods are needed to classify pulse signal. Pulse strength (PS) is the synthetical reflection of pulse force and its changing tread, and is hard to be represented by one or several characteristic parameters. Accordingly, the selection of feature points is more complicated. In this paper, a novel feature point extraction method was proposed for pulse waveform based on wavelet support vector regression (MWSVR). Support vector points were defined as pulse feature point. It cans not only representing the change of pulse signals at the same time also can reduced data sets via SVR. The experimental data was divided into three groups according to different surface pressure and each group has 4000 samples were used by training in literature. The experimental result shows that the proposed method is the validity and the usability in locating feature point of pulse signal.