【作者】 万龙；

【导师】 蔡坤宝；

【作者基本信息】 重庆大学 ， 信号与信息处理， 2010， 硕士

【摘要】 作为生物医学信号之一的脉象信号能反映人体脉搏的生理与病理信息,对于它的研究有助于人类加深对人体自身的认识,在防病与治病中发挥更大的作用,这是本文进行研究的目的。小波变换是一个迅速发展的新领域,它是在时域和频域都具有良好局域性的一种信号分析方法,而且非常适用于非平稳信号的处理。但是小波分析不适合非欧氏空间的应用,为了弥补传统小波的一些不足,提升小波应运而生,提升小波不仅具有通用、灵活的特点,而且还有高效的提升实现算法。由于有限长滤波器多相位矩阵的分解不是唯一的,而相同的多相位矩阵的提升实现算法不一定相同,本论文在总结前人工作的基础上,较深入地研究了应用提升小波变换对脉象信号进行特征提取与识别的方法,实现了多相位矩阵两种不同的分解格式及其相应的提升实现算法。其中一种分解格式及其相应的提升实现算法是由本文提出的.然后分别对40例脉象信号(20例健康正常人和20例海洛因吸毒者的脉象信号)进行一级提升小波变换,通过尺度系数,进行特征提取,找出了健康正常人与海洛因吸毒者的脉象信号之间的差异,初步提出了用于划分健康正常人与海洛因吸毒者的判据。本文的实验结果表明:第一种提升实现算法下18例健康正常人和19例海洛因吸毒者被检测出来,而健康正常人Z01和Z10被误判,海洛因吸毒者B13被误判;第二种提升实现算法下18例健康正常人和20例海洛因吸毒者被检测出来,而健康正常人Z01和Z10被误判。本论文还详细介绍了模糊C均值聚类的基本概念和理论依据。着重阐述了C均值模糊聚类的实现算法,在对脉象信号进行特征提取的基础上,利用该算法对20例健康正常人和20例海洛因吸毒者的脉象信号进行分类识别,初步聚类后的效果一般,通过标准差准则对误判的样本进一步划分后取得了不错的效果。更多还原

【Abstract】 As a part of the biomedicalsignals, Pulse signals can reflect the pulse of human physiological and pathological information. Studying it can help us to deepen our understanding on our body, and it could play a great role in disease prevention and treatment. It is the main research purpose in this paper.Wavelet transform is developing very qucikly in a new areas at present, it is a good signal analytical method both in the time and the frequency domains, especially applicable for non-stationary signal processing. But wavelet tranform is not suitable for non-European space’s application.In order to compensate for the shortcomings of the traditional wavelet,lifting wavelet generates.Lifting wavelet is not only common, flexible, but also it has effective lifting implementing algorithm. As the finite long filter’s decomposition of the multiphase matrix is not unique,and lifting implementing algorithm is not necessarily the same between the same multiphase matrix. Thus, based on previous work, this paper applys lifting wavelet transform to the pulse signals proceeding feature extraction and identification, achieves in two different multiphase matrix decomposition formats and the corresponding lifting implementing algorithm. One of the decomposition formats and the corresponding lifting implementing algorithm is proposed by this paper,then performs lifting wavelet transfom individually to the 40 pulse signals (20 healthy normal persons and 20 heroin addicts), through the scale coefficients, performing feature extraction, the difference between the healthy persons and the heroin addicts can be found,and a primary criterion for measuring the healthy normal persons and the heroin addicts can be obtained.There are 18 cases of the healthy normal persons and 19 cases of the heroin addicts that are detected in the first lifting algorithm, the healthy normal person Z01 and Z10 are misjudged, and the heroin addict B13 is misjudged, too.Also there are 18 cases of the healthy normal persons and 20 cases of the heroin addicts that are detected in the second lifting algorithm, the healthy person Z01 and Z10 are misjudged. Then the two methords are compared between the advantages and disadvantages.This paper also gives the basic concepts and theorise of the fuzzy C-means clustering in detail, and expounds the fuzzy C-means clustering algorithm. Based on the feature extraction of the pulse signals, this paper uses the algorithm to identify the pulse signals between th 20 healthy persons and the 20 heroin addicts. And there is a general result at the beginning, then the misjudged samples are devided by the standard deviation and a good result is achieved.更多还原