【作者】 刘丽丽；

【导师】 司玉娟；

【作者基本信息】 吉林大学 ， 电子与通信工程， 2012， 硕士

【摘要】 心电信号是一种非平稳微弱信号，主要含有工频干扰、基线漂移和肌电干扰噪声，去噪较困难。常用去噪算法主要包括小波变换、数字滤波器法、形态学滤波器等方法，存在计算复杂度较高或者滤波后信号失真等问题。自适应滤波器法可根据前一时刻己获得的滤波器参数等结果，实时调整滤波器参数，以适应信号和噪声未知的或随时间变化的统计特性。自适应滤波器无需输入信号的先验知识，适合低信噪比信号去噪，且滤波器结构简单，计算量小，处理信号实时性较高；依据LMS(Least Mean Square)准则，可使输出信号和输入信号的最小均方趋于相等，可有效避免消噪信号失真，最大限度的保留心电信号的特征信息。本文基于自适应滤波器对心电信号的去噪算法进行研究。本文选取双频级联陷波器去除工频干扰和基线漂移，二者均属于多频单色干扰。双频级联型陷波器由两个单频陷波器串联组成，不但保留了单频陷波器特性，而且参数调整相互独立。双频级联陷波器经过两级逐次迭代，收敛趋于稳定，可以有效去除工频干扰和基线漂移。心电信号属于周期信号，且肌电干扰是一种高频宽带信号，其频谱特性接近于高斯白噪声。鉴于以上特点，本文采用自适应信号分离器完成肌电干扰的滤除。自适应信号分离器去噪法具有针对性，它利用时间延时结构，分离宽带信号与周期信号，进而有效去除肌电干扰。心电信号信号突变点的检测是小波变换的重要应用之一。小波变换具有良好的时频分析特性，变换后系数的模值点或过零点通常对应于信号突变点。因此，本文对基于小波变换模极值法对心电信号QRS波群的检测算法进行研究。首先对心电信号进行3尺度分解，在第2、3尺度上通过寻找过零点定位R波，继而完成QRS波群起止点的检测。基于MATLAB软件平台对本文算法进行仿真实验，实验结果表明，基于自适应滤波器的ECG去噪算法能够较好的去除工频干扰、基线漂移和肌电干扰，具有较高的信噪比和较低的均方误差；基于小波变换的QRS识别算法，得到的识别率较高，可达98%以上。更多还原

【Abstract】 The ECG is a non-stationary random signal, containing mainly frequencyinterference, baseline drift and EMG interference noise, it is more difficult todenoising. Commonly denoising methods include wavelet transform, digital filtermethod, morphological filters, and other methods, computational complexity is highor the filtered signal distortion.Adaptive filter method according to the results of the first time have access to thefilter parameters, adjust the filter parameters to adapt to the unknown signal and noiseor statistical properties change on real-time. Adaptive filter does not require a prioriknowledge of the input signal, suitable for low SNR signal denoising filter structure issimple, small amount of calculation, the higher real-time signal processing; criteriabased on the Least Mean Square(LMS), can output the minimum mean square error ofthe signal and the input signal equal to each other, which can effectively avoid signaldistortion denoising to maximize the retention characteristics of the ECG information.ECG denoising algorithm based on adaptive filter.This paper selects the dual-band cascaded notch filter to remove frequencyinterference and baseline drift, both of which belong to the multi-frequencymonochromatic interference. Dual-band cascaded notch filter can retain thesingle-frequency notch filter characteristics, which consists of two single-band notchfilter in series, and the parameter adjustment independent of each other. Dual-bandcascaded notch filter after two successive iterations, convergence stabilized, canremove the frequency interference and baseline drift effectively.ECG signal is a periodic signal, and the EMG interference is a high frequencybroadband signals, their spectral characteristics close to Gaussian white noise. In lightof the above characteristics, this paper uses adaptive signal splitter to complete theEMG interference filter. Denoising method of adaptive signal separator targeted use ofthe time delay structure, separation of the broadband signal and periodic signal,thereby remove the EMG interference effectively.The detection of mutations of the ECG signal is one of the important applicationsof the wavelet transform. Wavelet transform has good time-frequency analysisfeatures, the point of the transform coefficients modulo value or zero generallycorresponds to the signal mutations. Therefore, based on wavelet transform modulusmaxima method of ECG QRS complex detection algorithm research. First ECG signalfor the three-scale decomposition, in paragraphs2,3scale by finding the zero positioning of R wave, then the beginning and ending points of the QRS complexdetection.This algorithm is based on MATLAB software platform simulation andexperimental results show better ECG denoising algorithm based on adaptive filter toremove frequency interference, baseline drift and EMG interference, high signal tonoise ratio and lower mean square error; QRS recognition algorithm based on wavelettransform, the higher recognition rate up to98%.更多还原