The Research On Signal Perception And Computer Aided Recognition Of Traditional Chinese Medicine Pulse Diagnosis

Computer-aided analysis of Traditional Chinese Medicine Pulse Diagnosis (TCMPD) involves several aspects such as the modernization of traditional Chinese medicine theory, sensor technology, design of sampling device and pattern recognition. Although some results are ready during past decades, that’s a long way to reach clinical requirements. From the points of view of pattern recognition, this dissertation pays attention to several key issues on computer-aided analysis of TCMPD.First of all, for there is no pulse wave database which meets demands on test and evaluation, the database containing detail labeled features and expert diagnostic results from single-point/multi-point, single part/multi-parts is presented.After that, the temporal features extraction with physiological significance is investigated. The extraction methods of the main wave based on Shannon energy envelope and Hilbert transform, the predicrotic wave and dicrotic wave based on the expertise threshold are presents.Late on, the deep convolutional neural network with9layers is used to recognize the pulse wave, and a noise module is added in the training process. The network construction is improved to apply other empirical features.Then, multi-point pulse wave sampling prototype device is designed, and the analysis to multi-point, multi-parts pulse wave is initialized. The output of TCMPD is inputted to Chinese Medicine Diet System, where signal sampling, analysis, diagnosis and health care could be integrated.Finally, further researches in TCPMD and time feature extraction are concluded. The physician’s experience combined with effective learning algorithm may be a possible ways to acquire valid feature for Computer-aided analysis of TCPMD.