基于CEEMD-VMD-SIST算法的sEMG信号降噪方法

实验研究表明,在基于表面肌电信号的手势识别中,由于噪声的存在,识别精度会大大降低；然而,传统的去噪方法由于对高频部分分解不当或模态混叠等问题暴露出缺点；针对传统表面肌电(sEMG)信号降噪方法对高频部分分解不当或频率混叠而导致降噪效果不佳,提出了一种基于互补集合经验模态分解(CEEMD)与变分模态分解(VMD)的滑动区间软阈值(SIST)降噪组合算法(CEEMD-VMD-SIST)；首先,通过CEEMD将含噪信号分解为从高频到低频的多个不同本征模态函数(IMF),并根据自相关系数客观界定信号的模态分量范围；然后,对选中的模态分量采用VMD的滑动区间软阈值方法进行分解降噪并与部分剩余模态分量进行重构；实验表明,在不同信噪比下,所提算法的降噪性能与传统降噪方法相比,信噪比与均方根误差均有明显改善,可以更大程度上保留信号的有用信息。

Denoising Method for sEMG Signal Based on CEEMD-VMD-SIST Algorithm

Experimental researches show that for gesture recognition based on surface electromyography (sEMG) signal, the recognition accuracy will be greatly reduced due to the presence of noise. However, for the traditional denoising methods, the shortcomings could be exposed by improper decomposition of high-frequency parts or mode mixing.Aiming at the poor denoising effect caused by traditional denoising methods with improper decomposition of high-frequency parts or frequency aliasing, an algorithm of sliding interval soft threshold (SIST) denoising based on complementary ensemble empirical mode decomposition and variational mode decomposition (CEEMD-VMD-SIST) is proposed for surface electromyography (sEMG) signal. First, the noisy signal is decomposed into a group of intrinsic mode functions (IMF) from high frequency to low frequency by CEEMD, and the modal component range of the signal is objectively determined according to the autocorrelation coefficients. Then, the components are decomposed and denoised by VMD-SIST and reconstructed with some remaining modal components. Experiments show that under different signal-to-noise ratios (SNR), the denoising performance of the proposed algorithm is significantly improved compared with the traditional denoising methods, and the two evaluation indices(SNR and RMSE(root mean square error)) are significantly improved, and the useful components of signal can be retained to a greater extent.