基于小波变换的心电信号去噪识别的研究与实现

发布时间：2018-07-12 11:09

本文选题：心电信号 + 平稳小波变换　；参考：《北京邮电大学》2017年硕士论文

【摘要】：心脏疾病是威胁人类生命的主要疾病,人们从未间断对此疾病的研究,随着患病人数的增多,心电信号的自动分析技术将成为未来的发展趋势。心脏疾病的自动化诊断是以心电信号的波形特征和一些信息为基础的,而这些特征和信息的获取是以纯净的心电信号为前提的。因此,为了促进心电信号自动分析技术的发展,加快心脏疾病预防、诊断和治疗的步伐,本文主要对心电信号自动分析技术中的预处理技术和波形检测技术进行研究,主要研究内容如下:1.利用实验室自主研发的心电信号采集装置采集到的心电数据和MIT/BIH心律失常数据库中的部分数据对本文提出的心电信号的预处理算法和波形检测算法进行验证。2.对心电信号的预处理,即噪声的去除进行了研究。以平稳小波变换和阈值去噪法为基础对心电信号中主要存在的基线漂移、工频干扰和肌电干扰这三种噪声进行滤除。首先,对常用的小波函数进行比较分析,最终选定小波函数bior5.5对心电信号进行8尺度的平稳小波分解;其次,为了克服软硬阈值函数在信号重构时出现的信号幅度失真、边缘模糊和伪吉布斯现象的缺陷,提出了一种新的阈值函数,并结合heursure阈值进行阈值去噪。实验结果表明,该种方法有效的去除了心电信号中的三种主要噪声,得到了较高的信噪比,保留了原始心电信号的基本波形特征。3.对心电信号的波形检测进行了研究。通过分析选择平滑函数的一阶导数即二次B样条小波作为小波函数,用其对心电信号进行4层小波分解。通过在第四尺度的细节系数cd4上对由R波所生成的模极值对进行检测,从而实现对R波的检测。在第二尺度的细节系数cd2上,以检测到的由R波所生成的模极值对为基础,在其左右两侧通过检测模极值对和过零点,进而实现对Q波、S波以及QRS波群起点与终点进行检测。针对Q波、R波、S波以及QRS波群起点与终点的检测提出了相应的防止漏检、防止误检和位置修正机制,进一步提高了波形检测的准确性。
[Abstract]:Heart disease is the main disease threatening human life. People have never stopped to study this disease. With the increase of the number of patients, the automatic analysis technology of ECG signal will become the development trend in the future. The automatic diagnosis of heart diseases is based on the waveform characteristics and some information of ECG signals, which are based on pure ECG signals. Therefore, in order to promote the development of ECG automatic analysis technology and accelerate the pace of heart disease prevention, diagnosis and treatment, this paper mainly studies the pretreatment technology and waveform detection technology of ECG automatic analysis technology. The main research contents are as follows: 1. The ECG data collected by the ECG acquisition device developed by the laboratory and part of the data in the MIT / BIH arrhythmia database are used to verify the ECG pretreatment algorithm and waveform detection algorithm proposed in this paper. The pretreatment of ECG signal, that is, noise removal, is studied. Based on stationary wavelet transform and threshold denoising method, three kinds of noise, namely baseline drift, power frequency interference and myoelectric interference, are filtered from ECG signals. Firstly, the wavelet function bior5.5 is chosen to decompose the ECG signal at 8 scales by comparing and analyzing the common wavelet function. Secondly, in order to overcome the amplitude distortion of the signal when the soft and hard threshold function occurs in the signal reconstruction, the wavelet function bior5.5 is selected to decompose the ECG signal at 8 scales. In this paper, a new threshold function is proposed for edge blur and pseudo-Gibbs phenomenon, and the threshold denoising is combined with heursure threshold. The experimental results show that the proposed method can effectively remove three kinds of main noises in ECG signal and obtain higher signal-to-noise ratio (SNR) and retain the basic waveform characteristics of the original ECG signal. 3. The waveform detection of ECG signal is studied. The first derivative of smoothing function, the quadratic B-spline wavelet, is selected as the wavelet function, and the wavelet decomposition of ECG signal is carried out. The detection of R wave is realized by detecting the mode extremum pair generated by R wave on the detail coefficient cd4 of the fourth scale. On the detail coefficient cd2 of the second scale, on the basis of the detected mode-extremum pair generated by R wave, we can detect the Q wave S wave and the starting and ending point of QRS wave group by detecting the mode extremum pair and the zero crossing point on the left and right sides. Aiming at the detection of Q wave / R wave / S wave and QRS wave group, this paper puts forward the corresponding mechanism to prevent missed detection, prevent false detection and position correction, and further improve the accuracy of waveform detection.
【学位授予单位】：北京邮电大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：R540.41;TN911.7

【相似文献】