The Objective Research For Auscultation In Traditional Chinese Medcine Based On Auditory Perception And Multi-classifiers Fusion

In traditional Chinese medicine and modern diagnostic engineering, auscultation is one of the diagnostic methods in traditional Chinese medicine (TCM). It obtains patients’ pathological changes through analysing patients’phonetic variation. Doctor of traditional Chinese medicine depends much on subjective hearing and easy to be effected by subjective factors and environment. This makes the diagnosis result lack of objectivity. Therefore, make sound diagnosis result is objective and quantitative change is particularly important.This paper studies the sound of traditional Chinese medicine diagnosis Xu Shi patients and normal speech feature extraction and recognition. In TCM, Xu Shi patients on sound performance:phonetic sound loud and clear voice resounding powerful, continuous person, more for Shi patients; Phonetic sound low thin, lazy to talk, the voice interrupted, before or after heavy light are Xu syndrome. So we can through the voice to discriminant is Xu syndrome or Shi syndrome. According to the Shanghai university of traditional Chinese medicine to provide Xu Shi patients and normal people’s/a:/sound samples, based on auditory physiological and psychological perception mechanism, the paper extract auditory perception coefficients, RASTA_PLP PLP, based on the equivalent bandwidth (ERB) extract the energy ratio and power ratio characteristics. Then using the characteristics of the proposed feature selection algorithm parameters selection. With pattern recognition classification algorithm of support vector machine (SVM) and integrated learning algorithm for Rotation Forest after the feature set classification, get52.07%and52.3%of the classification accuracy respectively.This thesis is focused on the diagnosis of five organ diseases and qi-deficiency, yin-deficiency,shi-zheng and normal people. According to the theory of TCM, there are5tones corresponding to5organs in human body. By listening to the changes in the5tones pronounced by the patient, the physician can determine the functional activity of the5organs.Voice signal perception process and the auditory system are inseparable. This thesis puts forward a method of using Gammatone filter bank to simulate basement membrane.This project utilizes the five tones and ten Chinese words provided by Shanghai University of Chinese Medicine. This paper extracts perceptual linear predicting parameters PLP, RASTA_PLP based on the physiological and psychological auditory perception mechanism, and extracts energy ratio and power ratio based on equivalent rectangular bandwidth (ERB) model. Later we conducted a feature selection procedure on the parameters we have extracted. It improve the efficiency of classification furthermore. At last, we have the classifying experiment on the optimized characteristics by applying the SVM and ensemble learning algorithm Rotation Forest, get80.83%and78.2%of the classification accuracy respectively.In addition, we extract vocal roughness, vocal brightness, sharpness of the same five tones and ten Chinese words based on vocal quality engineering principle. get55.43%and54.3%of the classification accuracy respectively.Finally, we have the primary discussion about vocal emotion recognition. We extracted the PLP, RASTA_PLP, energy ratio, power ratio and permutatio entropy of the samples in German emotional phonic database. get81.31%and82.22%of the classification accuracy respectively. We obtain some better classifying effect and discuss combining with the emotion theory in TCM tentatively.