The Pulse Signal Analysis And Research Based On HHT And SVM

Pulse diagnosis is one of the non-invasive methods of Traditional Chinese Medicine (TCM), But Traditional Chinese Medicine doctors usually make recipes according to their own experience, And pulse taking is difficult to inheritance. In order to improve the treatment performance, and to get the acknowledgement all over the world, it is urgent to make the TCM objective, standard, more accurate, and flexible to use in the clinical medical treatment.This paper puts forward an analysis method based on HHT and SVM for 8 kinds of pulse which include phase pulses. An improved filter to suppress the noise is proposed on the basis of EEMD and threshold method, the pulse signal de-noising processing seemed to be adaptive and effective. For de-noised pulse signals, to extract common time domain features from the waveform graph, and get new time and frequency domain features based on HHT.44 d features were selected by the progressive type of feature selection method. The two-step classification algorithm was improved, which combined with LIBSVM, to improve pulse condition signal pattern recognition research.The results show that the improved method can effectively detect and eliminate disturbances effect obviously The method that obviously improve the pulse signal de-noising effect greatly, the progressive type of feature selection method effectively filter the feature set, and select the more targeted, representative of the low dimensional characteristics in a relatively short time. In the study of classification algorithm, the improved two-step classification method is proposed to classify 482 pulse samples, Results show that the overall classification performance has been improved by 30%.