基于盲分离算法的心音信号分析

发布时间：2018-03-11 10:22

本文选题：心音　切入点：盲分离　出处：《华南理工大学》2012年硕士论文　论文类型：学位论文

【摘要】：心血管疾病是严重威胁人类健康的循环系统的疾病，心音信号是一种可以反映早期心脏病变的重要生理信号，尤其对瓣膜疾病的临床前诊断有着重要作用，对其的研究和分析受到越来越多学者的关注。临床上使用听诊器拾取心音，，通过医生的听觉系统被感知，依靠医生的经验做出诊断，容易受到主观因素的影响。在科研中常使用传感器摄取，借助放大电路等对心音信号进行调理，形成振动波形式的心音图，客观可见，便于保存，有助于进行深入分析心脏的生理状态。心脏正常的收缩舒张产生正常心音，对其进行研究，可以获得心率、明确心动周期，有利于对心脏的正常生理机能做出判断。心脏器质性病变等会产生心杂音。心杂音情况复杂，频率多变，振幅不一。传统的心音听诊难以准确区分心杂音和正常心音，心杂音混合在正常心音中，既干扰了对正常心音的分析，又不利于判断心脏病变。本文设计了一种基于LabView的时序同步双传感器系统，一传感器放置于心杂音的主要听诊区，获取包含心杂音成分的信息，另外一个传感器置于远离心杂音传导通路的心脏听诊区以降低杂音对正常心音的干扰，同步采集正常心音和心杂音权重不同的两路信号，作为两道参考信号输入盲分离算法网络，利用盲分析算法分离正常心音和心杂音，在此基础上对二者特征做单独分析。在源信号未知的情况下，盲分离技术可以解决从观测到的混迭信号中分离出源信号的问题，是信号处理领域研究的热点内容。本文首次将其应用于心音和心杂音的分离处理中。将正常心音形成系统作为一个振动源，而病理性心杂音产生的振动作为单独声源，在体表得到的心音图可以看做是两个源信号经心胸传导系统混叠之后的信号。以中心极限理论为主要依据，以峭度作为分离结果非高斯性质的衡量，采用梯度算法寻求最优解，对多例瓣膜病变信号进行了盲分离处理，重新构建心音的正常部分和杂音部分，并重点分析了二尖瓣狭窄和关闭不全两种病变导致的不同杂音，得到的时频特性可以很好的解释产生病变的原因，为实现瓣膜疾病的数字化诊断提供依据。
[Abstract]:Cardiovascular disease is a serious threat to human health of the circulatory system disease, the heart sound signal is a kind of early heart disease can reflect the important physiological signal, especially for valvular disease preclinical diagnosis has an important role. More and more scholars pay attention to its research and analysis. Clinical use of stethoscope to pick up heart sounds, sense through the doctor's auditory system, rely on the doctor's experience to make a diagnosis, easy to be affected by subjective factors. With the help of amplifying circuit, the heart sound signal can be adjusted to form the heart sound picture in the form of vibration wave, which can be seen objectively and is easy to preserve, which is helpful for the further analysis of the physiological state of the heart. The normal systolic and diastolic activity of the heart produces normal heart sounds, which can be studied to obtain the heart rate and determine the cardiac cycle. It is helpful to judge the normal physiological function of the heart. Heart diseases, such as organic diseases, produce cardiac murmur. The murmur of the heart is complex, the frequency is variable, and the amplitude is different. Traditional heart-sound auscultation is difficult to distinguish the murmur from the normal cardiac murmur accurately. The mixing of cardiac murmurs in normal heart sounds not only interferes with the analysis of normal heart sounds, but also does not help to judge heart disease. In this paper, a timing synchronous dual sensor system based on LabView is designed, one sensor is placed in the main auscultation area of heart murmur. The other sensor is placed in the heart-auscultation area away from the cardiac murmur conduction pathway to reduce the interference of the murmur to the normal heart sound, and simultaneously collect the two signals with different weights of normal heart sounds and cardiac murmurs. As the input of two reference signals, the blind analysis algorithm is used to separate normal heart sounds and heart murmur. When the source signal is unknown, the blind separation technique can solve the problem of separating the source signal from the observed mixed signal. It is a hot topic in the field of signal processing. In this paper, it is first applied to the separation of heart sounds and cardiac murmurs. The normal heart sound forming system is regarded as a vibration source, while the pathological heart murmur is used as a single sound source. The cardiogram obtained on the body surface can be regarded as the signal of two source signals mixed by the cardiothoracic conduction system. Based on the central limit theory and the kurtosis as the measure of the non-#china_person0# property of the separation result, the gradient algorithm is used to find the optimal solution. In this paper, the signals of several valvular lesions were separated, and the normal and murmur parts of the heart sound were reconstructed, and the different murmur caused by mitral stenosis and insufficiency were analyzed. The obtained time-frequency characteristics can explain the causes of the disease and provide the basis for digital diagnosis of valvular diseases.
【学位授予单位】：华南理工大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：R318.0

【参考文献】