基于两种改进阈值函数的表面肌电信号降噪研究

Research on SEMG signal denoising based on two improved threshold functions

针对传统小波阈值法在表面肌电信号去噪上存在的缺陷,提出逼近指数型阈值函数和逼近对数型去噪的方法。对含噪表面肌电信号进行改进阈值降噪处理。针对表面肌电信号的特点,在分析了传统的软阈值、硬阈值函数去噪原理的基础上,与软阈值函数、硬阈值函数和半软阈值在降噪效果上做比较。两种改进阈值函数克服了软阈值函数和硬阈值函数具有偏差性和不连续性及半软阈值函数具有偏差性及不连续性的缺点。通过选取给定阈值、固定阈值、估噪阈值三种阈值进行仿真实验,实验结果表明,在表面肌电信号去噪中两种改进阈值函数较传统软阈值函数和半软阈值函数具有较好的信噪比和较低的均方根误差。从阈值函数选取上看,硬阈值函数去噪效果最好,然后依次是指数型阈值函数、对数型阈值函数、线性阈值函数,最后是软阈值函数。从信噪比和均方根误差数值上看,给定阈值去噪效果最好,其次是固定阈值,最后是估噪阈值。

In view of the defect existing in SEMG signal denoising of the traditional wavelet threshold method,the approximating exponential type threshold function and the approximating logarithmic type denoising methods are proposed to improve the threshold denoising of SEMG signal with noise.According to the characteristics of SEMG,the denoising effect is compared with soft threshold function,hard threshold function and semi⁃soft threshold function on the basis of analyzing the denoising principle of traditional soft threshold and hard threshold functions.The two improved threshold functions overcome the defects of deviation and discontinuity of soft threshold function and hard threshold function,as well as the same defects of semi⁃soft threshold function.The given threshold value,fixed threshold value and threshold value of noise estimation are selected for the simulation experiments.The experiment results show that the two improved threshold functions have better signal⁃to⁃noise ratio and lower root⁃mean⁃square error than the traditional soft threshold function and semi⁃soft threshold function in SEMG signal denoising.Seen from the perspective of the threshold functions,the hard threshold function has the best denoising effect,followed by the exponential type threshold function,the logarithmic type threshold function,the linear threshold function and the soft threshold function in sequence.Seen from the signal⁃to⁃noise ratio and the root⁃mean⁃square error,the given threshold function has the best denoising effect,followed by the fixed threshold function,and finally the noise estimation threshold function.