The Objective Research For Auscultation In Traditional Chinese Medicine Based On Features Combination

Chinese auscultation, as one of the auscultation-olfaction in TCM four diagnostic methods, obtains patients’diagnostic results about pathology and disease location through analyzing patients’phonetic variation. Traditionally, this diagnostic method is based on the direct auditory sensitivity of doctors for the patients’pronunciation. It’s easy to be effected by the environment, doctors’experience and hearing and other factors, so it’s hard to avoid the misdiagnosis and missed diagnosis clinically. In order to overcome this difficulty, it’s necessary to change this subjective perceptual diagnosis into objective digitized diagnosis.This thesis studies the feature extraction and recognition of TCM five organ illness diagnosis. According to the description about the theory "five tones corresponding to five organs" in the classical works, five organs spleen, lung, liver, heart and kidney are corresponding to five tones gong, shang, jue, zhi, yu, the abnormality of five tones reveals the lesions occurred in five organs. Thus we can judge the variation of organs through analyzing the variation of the corresponding voice parameters. For five organs diseases sufferers and the normal people, features LSP, LPCC and ALPCC, MFCC and AMFCC and IMFCC, LPMFCC, MSE, SampEn, CB and so on are extracted. The experiments indicate, they can distinguish the six kinds of samples in the five organ illness diagnosis well at different levels.The degrees of features’diversity are found different through difference analysis. When we have difference analysis on each feature, it’s found that different features have different identification results for different groups. The thesis discusses the reasons why these features vary from different kinds of samples according to description in TCM five-organ-five-tone theory. We make the best of the complementary action brought by the difference to combine multiple features, getting the combined features set which can perform significant difference between any two groups. The classification accuracy on the features set finally reached to79.1%.This thesis also studies the TCM deficiency and excess syndromes experiment. The experiment classifies the qi deficiency syndrome, yin deficiency syndrome, excess syndrome and normal through looking for the complementary features by means of difference analysis. Finally, we also obtain good classification accuracy51.40%on the deficiency and excess syndromes experiment. Our study proves that the classification accuracy can be improved through the complementary features and it may be a good direction to promote the development of TCM auscultation. In addtion, this thesis develops a TCM-auscultation system preliminarily.