基于特征均值的SVD信号去噪算法

摘要

根据矩阵奇异值分解厚理,提出基于特征均值的信号去噪算法.该算法首先构造出加噪信号的Hankel矩阵,并对其进行SVD变换,再将小于全体特征值的均值的那些特征值置零,最后通过SVD反交换重建出去噪后的信号.通过与传统小波和FFT 信号去噪算法进行对比实验.结果表明,该方法具有较强的噪声鲁棒性,同时能更好地保留信号细节,但实现速度有所降低.%A signal denoising algorithm based on mean value of eigenvalue is proposed, which is according to the principle of matrix singular value decomposition. Firstly, this algorithm constructs a Hankel matrix with noised signal, and conducts SVD transformation on it; then it sets those eigenvalues to zero of which they are less than the mean value of all eigenvalues; finally it reconstructs the denoised signal through inverse SVD transformation. Experimental results of comparing the proposed method with traditional wavelet transform and FFT signal denois-ing method show that it has stronger noise robustness and can reserve signal details better, but its implementation speed is somewhat decreased.