表面肌电信号中工频干扰的滤波方法研究

Research on Filtering Methods for Power Frequency Interference in Surface Electromyography Signals

基于表面肌电信号的语音识别研究中,工频干扰会严重影响信号质量,使得识别准确率下降。如何在消除工频干扰的同时,尽可能保留肌电信号中的发音动作特征是一个值得研究的问题。在传统肌电信号的工频干扰滤除任务中,巴特沃斯带阻滤波器虽然消除了工频干扰,但也衰减了电信号中与工频干扰相同的频率成分。基于此,提出一种邻域均峰比滤波方法,动态计算频域滤波方法中的修正系数,针对工频干扰及其谐波进行修正,尽可能使被滤波的频带的幅度与相邻频带的幅度保持一致。对比结果表明,邻域均峰比滤波能够使滤波后的波形更接近原始波形。

In the research of speech recognition based on surface electromyography signals,power frequency interference can seriously affect signal quality,leading to a decrease in recognition accuracy.How to eliminate power frequency interference while preserving the pronunciation and movement characteristics of electromyographic signals as much as possible is a worthwhile research question.In the traditional task of filtering out power frequency interference in electromyography signals,although the Butterworth bandstop filter eliminates power frequency interference,it also attenuates the frequency components in the electrical signal that are the same as power frequency interference.Based on this,a neighborhood average to peak ratio filtering method is proposed,which dynamically calculates the correction coefficients in the frequency domain filtering method,corrects power frequency interference and its harmonics,and tries to make the amplitude of the filtered frequency band consistent with that of adjacent frequency bands as much as possible.The comparison results show that the neighborhood average to peak ratio filtering can make the filtered waveform closer to the original waveform.