A Novel Variable Step Size LMS Adaptive Filtering Algorithm

This paper proposes a novel variable step size LMS adaptive filtering algorithm. The algorithm is established based on nonlinear relationship between step and error signal in hyperbolic tangent function. The step is adjusted by the autocorrelation value of the error signal which is only influenced by the input signals. So this algorithm can accurately reflect the adaptive state and make the weight vector approach the best value. The algorithm not only performs faster in convergence speed and tracking speed, but also has better steady-state performance even in low signal to noise ratio (SNR) environment. Theoretical analysis and computer simulation show that this algorithm outperformed the other algorithms described in this article.