基于平移不变小波变换的脑电信号去噪

Denoising of EEG Signal Based on Translation Invariance Wavelet Transform

鉴于传统小波阈值法去噪时,硬阈值函数存在不连续性,软阈值函数存在恒定偏差,去噪效果不佳,提出一种基于新阈值函数的小波平移不变量去噪法。首先构造一种任意阶可导的新阈值函数,再采用平移不变小波方法处理脑电信号。在matlab2009平台上分别采用软、硬阈值及改进方法对脑电信号进行去噪,结果表明:改进方法能够更好的保持脑电信号的特征,且具有更高的信噪比和更低的均方根误差。该方法去除脑电信号噪声性能优于传统小波阈值法,有利于准确提取脑电特征参数。

A better denoising method of EEG signal based on wavelet transform is explored. The traditional wavelet denoising meth od is not very effective because the hardthreshold function is not continuous and the soft threshold function exists constant deviation. An improved wavelet threshold denoising method based on translation invariant was proposed. This method constructs a kind of new thresh old function which is arbitrary order derivable, then the EEG signal is processed by translation invariant method. EEG signal is denoised with the hardthreshold function, the softthreshold function and the improved method respectively. Experimental results indicate that this method is better than traditional wavelet thresholding denoising methods in aspects of remaining characteristics of EEG signal and has higher SNR and lower RMSE. The proposed method is better than traditional wavelet method in EEG signal denoising, and is conducive to extract the precise feature of EEG.