心电波形聚类策略的改进方法研究

发布时间：2018-02-26 16:37

本文关键词： 心电信号聚类小波变换 K-means　出处：《哈尔滨理工大学》2015年硕士论文　论文类型：学位论文

【摘要】：心电图（Electrocardiogram，ECG）能够直观的展现心脏兴奋的电活动过程，是人类生命体征的一个重要表现。医学中，针对心电波形信号的诊断成为医生在检测心脏疾病的重要一环。目前的心电波形信号自动分析方法还存在着许多的缺点和不成熟的地方，无论是在理论研究还是实际应用方面，仍有许多需要进行改进和创新的地方。为了改善这一现状，论文围绕：“心电波形信号的预处理”、“心电波形信号的检测与定位”、“心电波形信号的聚类”三个方面的关键技术展开研究。首先，进行对心电波形信号的预处理。为了减少心电信号中的工频干扰和基线漂移等干扰噪声，根据心电信号自动分析的实际需要，对不同尺度的小波使用相应的阈值处理小波系数，解决了小波阈值去噪法最优分解层数难以确定的问题。经过MIT-BIH中心电数据实验验证，，改进后的小波滤波器既能滤除工频干扰和高频噪声，又能较好的保留有用信号细节。其次，本文应用小波变换实现QRS波形检测，改进心电的传统小波变换方法，使其克服计算复杂，检测速度较慢，对伪差信号容易误判等缺点。针对心电波形信号的P波、T波定位与检测困难的问题，应用自适应对消法的思想，进行QRS-T波形自适应对消，增大P波在剩余波形中所占比例。该P波检测方法能够较好的排除高频噪声和基线漂移的影响，并且QRS波形的形态以及出现心律失常等情况并不会影响此检测方法，具有较好的P波检测效果。最后，为了改进心电波形信号的自动聚类，提出了一种改进聚类算法：基于均方差属性加权的遗传模拟退火K-means改进聚类算法，对心电信号进行波形聚类。采用MIT-BIH标准心电数据库中的典型心电数据，对算法进行实验判断，比较本文改进后的K-means聚类算法与传统的K-means聚类算法的聚类结果，本文算法聚类准确率胜出，实验证明本算法能够对心电信号有效聚类。
[Abstract]:Electrocardiogram electrocardiogramme (ECG) can directly show the electrical activity of the heart, which is an important manifestation of human vital signs. The diagnosis of ECG waveform signals has become an important part of doctors' detection of heart disease. There are still many shortcomings and immaturity in the current automatic analysis methods of ECG waveform signals, both in theoretical research and in practical applications. There is still much room for improvement and innovation. This paper focuses on three key technologies: "pretreatment of ECG waveform signal", "detection and location of ECG waveform signal" and "clustering of ECG waveform signal". First of all, the ECG waveform signal is preprocessed. In order to reduce the interference noise such as power frequency interference and baseline drift in ECG signal, according to the actual need of ECG automatic analysis, The wavelet coefficients are processed by the corresponding threshold for different scales, which solves the problem that the optimal decomposition layer number of wavelet threshold denoising method is difficult to determine. The improved wavelet filter can not only filter power frequency interference and high frequency noise, but also retain useful signal details. Secondly, wavelet transform is used to realize QRS waveform detection, and the traditional wavelet transform method of ECG is improved to overcome the complexity of calculation and slow detection speed. In view of the difficulty of P wave T wave localization and detection of ECG waveform signal, the idea of adaptive pair cancellation method is applied to carry out adaptive cancellation of QRS-T waveform. The P wave detection method can eliminate the influence of high frequency noise and baseline drift, and the shape of QRS waveform and the occurrence of arrhythmia will not affect this detection method. It has better P wave detection effect. Finally, in order to improve the automatic clustering of ECG waveform signals, an improved clustering algorithm is proposed: genetic simulated annealing K-means clustering algorithm based on weighted mean square error attribute. Using the typical ECG data in MIT-BIH standard ECG database, the algorithm is judged experimentally, and the clustering results of the improved K-means clustering algorithm and the traditional K-means clustering algorithm are compared. The experimental results show that the algorithm can effectively cluster ECG signals.
【学位授予单位】：哈尔滨理工大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TN911.7;R318.04

【参考文献】