摘要
记录舒张期心音信号,用信号处理方法进行分析识别,实现冠心病的无创诊断.将小波分析(WT)、自回归滑动平均(ARMA)以及独立分量分析(ICA)方法分别应用于冠心病(CAD)心音信号的特征提取,并将提取的特征值输入径向基(RBF)神经网络进行训练和识别.实验结果,CAD病人和非CAD病人的正确检测率分别是:小波分析80%,85%;ARMA70%,75%;ICA85%,85%.结果表明,在CAD病人的心音中含有300～800Hz的高频心音能量.在三种方法中,ICA显示了较好的效果.
The mechanism of coronary artery blood indicates that heart sounds contain the information of coronary artery disease (CAD). The blood flow diastole in narrow artery causes laminar current or turbulency. Through efficient signal processing, the parameters of the diastolic blood flow in the coronary artery was extracted to diagnose CAD noninvasively. This research applied wavelet analysis (WT), auto-regressive and moving average (ARMA) and independent component analysis (ICA) to heart sounds obtained by sensor array to extract the characteristics of the heart sounds. These characteristic parameters were inputted to radial basis function networks (RBF) for identification. The results show that the diagnosis correctness is 82.5% by wavelet analysis, 72.5% with ARMA and 85% by ICA.
出处
《浙江大学学报（工学版）》
EI
CAS
CSCD
北大核心
2004年第1期98-102,共5页
Journal of Zhejiang University：Engineering Science
基金
浙江省重点科研计划资助项目(971103099).
关键词
心音信号
冠心病
小波分析
自回归滑动平均
独立分量分析
径向基神经网络
Acoustic signal processing
Biosensors
Diseases
Independent component analysis
Laminar flow
Radial basis function networks
Regression analysis
Turbulent flow
Wavelet transforms
