Title: Nonlinear analysis of auscultation signals in Traditional Chinese Medicine using Wavelet Packet Transform and Approximate Entropy

Authors: Jianjun Yan, Yong Shen, Yiqin Wang, Fufeng Li, Chunming Xia, Rui Guo, Chunfeng Chen, Zhongyan Gu, Xiaojing Shen

Addresses: Center for Mechatronics Engineering, East China University of Science and Technology, Shanghai, PR China. ' Center for Mechatronics Engineering, East China University of Science and Technology, Shanghai, PR China. ' Center for TCM Information Science and Technology, Shanghai University of TCM, Shanghai, PR China. ' Center for TCM Information Science and Technology, Shanghai University of TCM, Shanghai, PR China. ' Center for Mechatronics Engineering, East China University of Science and Technology, Shanghai, PR China. ' Center for TCM Information Science and Technology, Shanghai University of TCM, Shanghai, PR China. ' Center for TCM Information Science and Technology, Shanghai University of TCM, Shanghai, PR China. ' Center for Mechatronics Engineering, East China University of Science and Technology, Shanghai, PR China. ' Center for Mechatronics Engineering, East China University of Science and Technology, Shanghai, PR China

Abstract: The distinctive characteristic of Wavelet Transform (WT) is that it can well characterise the local information of signals in time-frequency domain, and Wavelet Packet Transform (WPT) has a more subtle decomposition method than WT. The purpose of this paper is to analyse the auscultation signals in Traditional Chinese Medicine (TCM) utilising WPT and Approximate Entropy (ApEn). In this paper, a new scheme was presented for analysing the Auscultation Signals consisted of qi-deficient, yin-deficient and normal people. In the first stage, voice signal were decomposed into approximation and detail coefficients using WPT. Then the ApEn values of these signals were computed based on these coefficients. The differences of the ApEn values and the meaning of which for signals among three kinds of samples were discussed. Finally the conclusion can be drawn that the distributions of ApEn in different frequency ranges for all signals of three kinds of samples have their special characteristics. The ApEn values for three kinds of samples were used as the feature vectors for Support Vector Machine (SVM) classifier and some impressing results of classifications can be obtained.

Keywords: auscultation signals; WPT; wavelet packet transform; nonlinear analysis; approximate entropy; SVM; support vector machines; TCM; traditional Chinese medicine; China; wavelet transform; voice signals; frequency ranges; classification.

DOI: 10.1504/IJFIPM.2009.030831

International Journal of Functional Informatics and Personalised Medicine, 2009 Vol.2 No.3, pp.325 - 340

Published online: 07 Jan 2010 *