小波包分析和FastICA相结合对单通道脑电信号的去噪研究

Study on monopolar-channel EEG signal denoising by wavelet packet analysis combined with FastICA

采用基于小波包分析的FastICA方法对单通道脑电信号进行去噪,由于脑电信号并不是独立的,分离信号和源信号对应关系也不确定,直接采用FastICA对多通道脑电信号进行处理,最终的去噪效果并不理想。因此先用小波包对单通道脑电信号进行分解,从而得到频带较窄的子带信号,可保证源信号具有一定的独立性,有利于噪声独立分量的分离和去除,也为单通道脑电信号的去噪提供了思路。从时域分析、傅里叶变换、功率谱密度分析以及信噪比和均方根误差的角度,对比多导联FastICA的去噪效果,可以发现所提出的方法较为有效。

The monopolar channel EEG signal is denoised with FastICA method based on wavelet packet analysis.For multi-channel EEG signals are not independent,and the corresponding relationship between the separated signals and the source signals is uncertain,it is not effective to denoise multi-channel EEG signals by FastICA alone.Therefore,a wavelet packet is used to decompose the monopolar-channel EEG signal first to obtain the sub-band signal with narrower frequency band,which can ensure that the source signals have a certain independence,so as to facilitate the separation and removal of independent components of noise.Therefore,it provides an idea for the denoising of monopolar-channel EEG signals.In the aspects of time domain analysis,Fourier transform,power spectral density analysis,signal-to-noise ratio(SNR)and root-mean-square error(RMSD),the denoising effect of multi-channel FastICA is compared with that of the method proposed in this study,from which it is found that the proposed method is more effective.