Design Of The Classification System And Study Of Recognition Based On ANN For The Manifestations Of The Pulse In Chinese Medicine

Data acquisition and pattern recognition of manifestations of the pulse in Chinese medicine are studied and discussed in this thesis on the basis of the research and development of AI recognition system for manifestations of the pulse in Chinese medicine.Firstly, based on the characteristics of manifestations of the pulse in Chinese medicine, a data acquisition circuit is designed. In addition, the system function of time domain feature extraction and pattern recognition of manifestations of the pulse is primarily implemented. Secondly, a multilayered neural network trained with a learning vector quantization ( LVQ ) algorithm is applied to pattern recognition of manifestations of the pulse and the classification ability of LVQ network is compared with traditional near neighbor algorithm.Moreover, the influences that number and selection methods of initializing quantization vectors have over the classification ability are discussed. Then a modified LVQ algorithm is presented, and the classification ability of network is improved further through selecting proper parameters.In spite of limited training samples, the simulation results show that ANN is a feasible and effective way to pattern recognition of manifestations of the pulse in Chinese medicine. It is superior to traditional pattern recognition methods as far as self-adaptive ability is concerned.