摘 要:针对心音的特征提取问题,将一维心音信号转换成二维心音图,采用图#br# 像处理技术提取心音的图像特征。本文首先对一维心音信号进行小波降噪和幅值归一化,将#br# 处理后的心音信号转换成具有统一性和可比性的二维心音图,并进行预处理;然后结合心音#br# 生理意义和二维心音图的图像特征,对能表征二维心音图生理信息的图像特征进行分析研#br# 究,重点研究二维心音图纵横坐标比和拐点序列码特征;最后,基于纵横坐标比、拐点序列#br# 码、小波分解系数3 个特征,探讨利用欧氏距离和支持向量机(SVM)两种识别方法进行#br# 二维心音图分类和身份识别的可行性。实验结果表明,3 种特征都可以实现二维心音图的分#br# 类识别,其中拐点序列码识别率最高;这种基于图像处理的二维心音图分类和身份识别方法#br# 具有明显的可行性和实用性,拥有广阔的应用前景。#br# 关 键 词:二维心音图;图像处理 ;特征提取;识别
Abstract: The one-dimensional heart sound signal is converted into a two-dimensional#br# phonocardiogram, then image feature of heart sounds based on image processing technology in a#br# two-dimensional phonocardiogram is extracted. Firstly the wavelet noise reduction and amplitude#br# normalization of one-dimensional heart sound by one-dimensional signal processing method are#br# realized, and then heart sounds after the treatment are converted into two-dimensional#br# phonocardiogram with uniformity and comparability, and pretreatment. And the image#br# characteristics of two-dimensional phonocardiogram are analyzed, which is characterization of#br# heart sounds’ physiological information combining with heart sounds’ physiological significance#br# and two-dimensional phonocardiogram’s image features, and the focus is on vertical and#br# horizontal ratio of coordinate and sequence code of inflection point. At last, the feasibility of#br# classification and identification is explored of 2D-PCG using Euclidean distance and Support#br# Vector Machine (SVM) based on vertical and horizontal ratio of coordinate, sequence code of#br# inflection point and wavelet coefficients. Experimental results show that the three features can#br# achieve the classification and recognition of the two-dimensional phonocardiogram, and inflection#br# point sequence code gets the highest recognition rate. The method of 2D-PCG classification and#br# identification based on a two- image processing has the feasibility and practical applicability, and#br# has broad application prospects.#br# Key words: two-dimensional phonocardiogram; image processing; feature extraction;#br# recognition
摘要: 针对心音的特征提取问题,将一维心音信号转换成二维心音图,采用图
像处理技术提取心音的图像特征。本文首先对一维心音信号进行小波降噪和幅值归一化,将
处理后的心音信号转换成具有统一性和可比性的二维心音图,并进行预处理;然后结合心音
生理意义和二维心音图的图像特征,对能表征二维心音图生理信息的图像特征进行分析研
究,重点研究二维心音图纵横坐标比和拐点序列码特征;最后,基于纵横坐标比、拐点序列
码、小波分解系数3 个特征,探讨利用欧氏距离和支持向量机(SVM)两种识别方法进行
二维心音图分类和身份识别的可行性。实验结果表明,3 种特征都可以实现二维心音图的分
类识别,其中拐点序列码识别率最高;这种基于图像处理的二维心音图分类和身份识别方法
具有明显的可行性和实用性,拥有广阔的应用前景。