基于心电信号循环平稳特征的心脏性猝死识别研究

发布时间：2018-05-30 12:54

本文选题：心电信号 + 心电信号建模　；参考：《兰州理工大学》2017年硕士论文

【摘要】：心脏性猝死是一种对人类生命有巨大威胁的疾病,大多数学者将猝死的时间限定在发病1小时之内,如果能在发病前对心脏性猝死疾病进行预警,就可以在心脏性猝死发生前挽救患者的生命。心电信号是心脏机械收缩与舒张的反映,是非平稳信号,但却表现出一定的准周期特性,即心电信号具有循环平稳特性,当人体内发生心脏性猝死疾病时,自身的循环平稳特性也会发生改变。传统的心血管疾病识别是把心电信号作为平稳信号进行分析,但心电信号是非平稳的,因此采用循环平稳算法对心电信号进行特征提取,结合支持向量机对心脏性猝死进行识别。论文主要工作如下:(1)心电信号滤波是特征提取的前提,针对心电信号采集过程中的常见噪声,采用适合非平稳信号处理的小波变换算法对心电信号进行滤波。首先对心电信号进行建模仿真,产生干净的心电信号,通过添加不同信噪比的噪声来评价滤波器效果。其次设计了小波变换滤波器,通过和整系数滤波器相比较,本文设计的小波变换滤波器对心电信号的滤波效果更好。最后用实际心电信号对小波变换滤波器进行验证,实验结果表明:小波变换滤波能够有效地去除高频和低频噪声。(2)根据心电信号表现出的循环平稳特性,首先介绍了循环平稳的基本概念一阶和二阶循环平稳,并由二阶循环平稳推导了反映心电信号循环平稳特性的积分循环功率谱密度函数;接着详细分析了实时性较高的时域平滑循环谱估计算法——FFT累加算法,并对正弦信号进行了循环谱的理论计算和仿真估计,实验结果表明:循环谱仿真估计结果与理论计算结果一致;最后在此基础上,提取不同典型人群心电信号的循环平稳特征,并利用心电信号的循环平稳特性对干扰段进行了检测。(3)针对心脏性猝死识别准确率不高的问题,提出了基于心电信号循环平稳特征的识别方法。在循环平稳理论的基础上,提取了不同典型人群心电信号的循环平稳特征,结合支持向量机对心脏性猝死进行识别,实验得出循环频率均值最能反映循环平稳特征,比较了两类线性分类器与支持向量机的识别效果,最后采用支持向量机和现有心脏性猝死识别方法进行对比,实验结果表明基于心电信号循环平稳特征的心脏性猝死识别方法在准确性上有明显的优势,猝死心电信号的识别准确率最高可达97.50%。
[Abstract]:Sudden cardiac death is a disease with great threat to human life. Most scholars limit the time of sudden death to one hour. The patient's life can be saved before sudden cardiac death occurs. ECG signal is the reflection of cardiac mechanical contraction and relaxation, which is non-stationary signal, but it shows certain quasi-periodic characteristic, that is, electrocardiogram signal has the characteristic of circulatory stability, when sudden cardiac death occurs in human body, Its own cycle stability will also change. In traditional cardiovascular disease recognition, ECG signals are analyzed as stationary signals, but ECG signals are non-stationary. Therefore, a cyclic stationary algorithm is used to extract the features of ECG signals. To identify sudden cardiac death with support vector machine (SVM). The main work of this paper is as follows: (1) ECG filtering is the premise of feature extraction. Aiming at the common noise in ECG signal acquisition, wavelet transform algorithm suitable for non-stationary signal processing is used to filter ECG signal. Firstly, the ECG signal is modeled and simulated to generate clean ECG signal, and the filter effect is evaluated by adding different SNR noise. Secondly, the wavelet transform filter is designed. Compared with the integer coefficient filter, the wavelet transform filter designed in this paper has better effect on ECG signal filtering. Finally, the actual ECG signal is used to verify the wavelet transform filter. The experimental results show that the wavelet transform filter can effectively remove the high frequency and low frequency noise. Firstly, the basic concepts of cyclic stationarity, first and second order, are introduced, and the integral cyclic power spectral density function, which reflects the cyclic stationary characteristic of ECG signal, is derived from the second-order cyclic stationarity. Then, the time-domain smoothing cyclic spectrum estimation algorithm, FFT accumulative algorithm, is analyzed in detail, and the theoretical calculation and simulation of the cyclic spectrum of sinusoidal signal are carried out. The experimental results show that the simulation results of cyclic spectrum are consistent with the theoretical results. Finally, the cyclic stationary characteristics of ECG signals of different typical populations are extracted. In order to solve the problem that the accuracy of sudden cardiac death recognition is not high, a method based on cyclic stationary characteristic of ECG signal is proposed to detect the disturbance segment. On the basis of the theory of cyclic stationary, the cyclic stationary characteristics of ECG signals of different typical people are extracted, and the sudden cardiac death is identified by using support vector machine. The experimental results show that the mean value of cycle frequency can best reflect the characteristics of circulatory stability. The recognition effects of two kinds of linear classifiers and support vector machines (SVM) are compared. Finally, support vector machines (SVM) and existing methods of sudden cardiac death (SCD) recognition are compared. The experimental results show that the recognition method of sudden cardiac death based on the steady characteristics of ECG cycle has obvious advantages in accuracy, and the recognition accuracy of sudden death signal can reach 97.50%.
【学位授予单位】：兰州理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：R541.78;TN911.7

【参考文献】