心电信号自动分析技术研究
发布时间：2018-10-23 20:42
【摘要】：心电图一直是人们了解自身心脏特征的主要途径，是疾病诊断的重要依据。由于动态心电图的产生，导致手工分析心电图所有数据已经不可能，为了提高诊断效率，实时监测病人，心电信号自动分析技术的诞生是必然的。 心电信号属于微弱信号，采集信号一般都包含各种干扰噪声。因此，首先要对信号进行去噪处理，心电去噪是QRS波形检测和特征提取的基础，其结果将直接影响到自动分析的诊断结果。QRS波是心电图中最明显的部位，包含了很多重要生理信息，因此，QRS检测是自动分析中重要的一步，不仅是其他波形定位的基础，也是特征提取的前提，将影响自动分析诊断的准确度。本文基于前人研究成果的基础上，主要对QRS波群检测技术做了进一步研究。 在去噪处理方面，本文采用了小波阈值去噪法。主要工作：1.选择合适的小波函数并确定小波分解层数。2.选择合适的阈值函数和阈值估计方法。3.进行仿真实验。本文利用sym8小波对心电进行小波分解，采用小波硬阈值法处理信号，同时通过输出信噪比(SNR)和最小均方误差(MSE)两参数对实验结果进行评估，表明该方法能够有效的去除心电信号中的主要噪声，具有很好的去噪效果。 本文提出了一种基于小波变换的R峰定位方法。该方法采用了高斯小波作为小波函数，选取了能量集中、噪声等干扰较弱的第3层的小波分解系数作为研究对象。主要工作：1.初始阈值并确定自动阈值变换规则。2.寻找在符合阈值条件的极值点，并通过一定方法以及优化策略对极值点进行正确配对。3.根据极值对确定R波峰在原信号的位置区间，并在该区间中找出最值，该值位置即为R波峰位置。4.通过不应期生理原理对R定位结果进行误检。5.仿真实验。本文对MIT-BIH数据库中部分典型波形数据进行了实验，实验结果表明该算法定位R波峰的正确率很高，是一个有效性算法。 以定位R波峰为基础，本文实现了QRS波群宽度的提取。主要工作：1.确定Q、S波在R波旁边的大概范围，在此范围中寻找极值点，，并对这些极值点进行正确配对2.根据极值对确定R波峰在原信号的位置区间，并在该区间中找出最值，该值位置即为波峰（Q或S波峰）。3.在Q波峰前或S波峰后8个采样点中寻找斜率变化最大的采样点，并认为该点为Q波起始点或S波终点，即为QRS波群的始点和终点，算出QRS波群宽度5.仿真实验，在波形中标出QRS波群宽度、Q波峰及S波峰。通过对MIT-BIH数据库中部分典型波形数据进行了实验，实验结果表明该方法有较好的精确度。
[Abstract]:Electrocardiogram (ECG) is the main way for people to understand their heart characteristics and an important basis for disease diagnosis. Because of the production of dynamic electrocardiogram, it is impossible to analyze all ECG data manually. In order to improve diagnosis efficiency and monitor patients in real time, the birth of ECG automatic analysis technology is inevitable. ECG signals belong to weak signals, and the collected signals generally contain all kinds of interference noise. Therefore, first of all, the signal should be de-noised. ECG denoising is the basis of QRS waveform detection and feature extraction, and the results will directly affect the diagnosis result of automatic analysis. QRS wave is the most obvious part of ECG. Therefore, QRS detection is an important step in automatic analysis, which is not only the basis of other waveform localization, but also the premise of feature extraction, which will affect the accuracy of automatic analysis and diagnosis. Based on the previous research results, this paper mainly focuses on the QRS wave group detection technology. In the aspect of denoising, wavelet threshold denoising method is adopted in this paper. Main work: 1. Select the appropriate wavelet function and determine the number of wavelet decomposition layers. 2. Select appropriate threshold function and threshold estimation method. 3. The simulation experiment is carried out. In this paper, sym8 wavelet is used to decompose ECG, and wavelet hard threshold method is used to process the signal. At the same time, the experimental results are evaluated by output SNR (SNR) and minimum mean square error (MSE). It shows that this method can effectively remove the main noise in ECG signal and has a good denoising effect. In this paper, a method of R peak location based on wavelet transform is proposed. In this method, Gao Si wavelet is used as the wavelet function, and the wavelet decomposition coefficient of the third layer, where the energy concentration and noise are weak, is chosen as the object of study. Main work: 1. Initial threshold and determine automatic threshold transform rules. 2. To find the extremum that meets the threshold condition, and make the correct pairing of the extremum by certain method and optimization strategy. 3. According to the extreme value pair, the position interval of R wave peak in the original signal is determined, and the maximum value is found in the interval. The position of the value is the position of the R wave peak. 4. The results of R localization were detected by the physiological principle of refractory period. 5. 5. Simulation experiment. In this paper, some typical waveform data in MIT-BIH database are experimented. The experimental results show that the algorithm has a high accuracy and is an effective algorithm. On the basis of locating R wave peak, the width of QRS wave group is extracted in this paper. Main work: 1. Determine the approximate range of QS waves next to R waves, search for extremum points in this range, and make the correct matching of these extreme points 2. According to the extreme value pair, the position interval of R wave peak in the original signal is determined, and the maximum value is found in the region. The position of the value is the peak (Q or S wave peak). 3. Among the 8 sampling points in front of Q wave peak or after S wave peak, the sampling point with the greatest slope change is found. It is considered as the starting point of Q wave or the end point of S wave, that is, the beginning and end point of QRS wave group, and the width of QRS wave group is calculated. Simulation experiments show that the width of QRS wave group, Q wave peak and S wave peak are marked in the waveform. Experiments on some typical waveform data in MIT-BIH database show that this method has good accuracy.
【学位授予单位】：南昌大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TN911.6
本文编号：2290410
[Abstract]:Electrocardiogram (ECG) is the main way for people to understand their heart characteristics and an important basis for disease diagnosis. Because of the production of dynamic electrocardiogram, it is impossible to analyze all ECG data manually. In order to improve diagnosis efficiency and monitor patients in real time, the birth of ECG automatic analysis technology is inevitable. ECG signals belong to weak signals, and the collected signals generally contain all kinds of interference noise. Therefore, first of all, the signal should be de-noised. ECG denoising is the basis of QRS waveform detection and feature extraction, and the results will directly affect the diagnosis result of automatic analysis. QRS wave is the most obvious part of ECG. Therefore, QRS detection is an important step in automatic analysis, which is not only the basis of other waveform localization, but also the premise of feature extraction, which will affect the accuracy of automatic analysis and diagnosis. Based on the previous research results, this paper mainly focuses on the QRS wave group detection technology. In the aspect of denoising, wavelet threshold denoising method is adopted in this paper. Main work: 1. Select the appropriate wavelet function and determine the number of wavelet decomposition layers. 2. Select appropriate threshold function and threshold estimation method. 3. The simulation experiment is carried out. In this paper, sym8 wavelet is used to decompose ECG, and wavelet hard threshold method is used to process the signal. At the same time, the experimental results are evaluated by output SNR (SNR) and minimum mean square error (MSE). It shows that this method can effectively remove the main noise in ECG signal and has a good denoising effect. In this paper, a method of R peak location based on wavelet transform is proposed. In this method, Gao Si wavelet is used as the wavelet function, and the wavelet decomposition coefficient of the third layer, where the energy concentration and noise are weak, is chosen as the object of study. Main work: 1. Initial threshold and determine automatic threshold transform rules. 2. To find the extremum that meets the threshold condition, and make the correct pairing of the extremum by certain method and optimization strategy. 3. According to the extreme value pair, the position interval of R wave peak in the original signal is determined, and the maximum value is found in the interval. The position of the value is the position of the R wave peak. 4. The results of R localization were detected by the physiological principle of refractory period. 5. 5. Simulation experiment. In this paper, some typical waveform data in MIT-BIH database are experimented. The experimental results show that the algorithm has a high accuracy and is an effective algorithm. On the basis of locating R wave peak, the width of QRS wave group is extracted in this paper. Main work: 1. Determine the approximate range of QS waves next to R waves, search for extremum points in this range, and make the correct matching of these extreme points 2. According to the extreme value pair, the position interval of R wave peak in the original signal is determined, and the maximum value is found in the region. The position of the value is the peak (Q or S wave peak). 3. Among the 8 sampling points in front of Q wave peak or after S wave peak, the sampling point with the greatest slope change is found. It is considered as the starting point of Q wave or the end point of S wave, that is, the beginning and end point of QRS wave group, and the width of QRS wave group is calculated. Simulation experiments show that the width of QRS wave group, Q wave peak and S wave peak are marked in the waveform. Experiments on some typical waveform data in MIT-BIH database show that this method has good accuracy.
【学位授予单位】：南昌大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TN911.6
【参考文献】
相关期刊论文 前10条
1 戚建新，卞正中，杨强;基于微机的心电信号实时自动分析系统[J];北京生物医学工程;1997年03期
2 金怡果,刘艳斌;一种检测与分析心电波形的算法[J];福州大学学报(自然科学版);1998年01期
3 刘希顺,王博亮;基于小波变换的心电图信号中QRS综合波检测算法[J];国防科技大学学报;1997年05期
4 苏丽,赵国良,李东明;心电信号QRS波群检测算法研究[J];哈尔滨工程大学学报;2005年04期
5 王伟;李小梅;;利用自适应小波变换检测心电QRS波算法的实现[J];科技通报;2008年01期
6 戴建新;郦志新;宋洪雪;;基于小波的信号Lipschitz指数分析和应用[J];南京邮电大学学报(自然科学版);2008年06期
7 余辉,张力新,吕扬生;基于小波变换的QRS波检测[J];生物医学工程与临床;2001年02期
8 夏恒超,詹永麒,杨海威;基于小波变换和移动窗口积分函数的心电信号的QRS波起、终点的检测[J];上海交通大学学报;2002年12期
9 曹玉珍,丁北生,吕扬生,高贞;小波变换用于微分心电信号突变点检测[J];天津大学学报;2001年05期
10 陈作炳,钱湘萍,李春光;用小波变换对心电图进行分析[J];武汉理工大学学报;2001年05期
相关博士学位论文 前2条
1 孙轶;基于自适应提升小波的信号去噪技术研究[D];中国科学技术大学;2008年
2 姚成;心电信号智能分析关键技术研究[D];吉林大学;2012年
本文编号：2290410
本文链接：https://www.wllwen.com/yixuelunwen/swyx/2290410.html
