A Variable-Step LMS-Based Adaptive Filtering Algorithm Using Standard Deviation

Adaptive filters are used in digital signal processing for a variety of applications including echo cancellation, noise suppression, channel equalization, source separation, and channel estimation. Their main function is to automatically and dynamically adjust their coefficients to adapt to the changing characteristics of the input signal. In this work it is presented a new adaptive filtering algorithm that is aimed at reducing the estimation error and improving the speed of convergence with respect to existing algorithms. The proposed algorithm is variable step and adjusts the coefficients of the digital filter based on the standard deviation of the estimation error and its derivative. This is a new approach for the adaptation of the coefficients that had not been explored until now. The results obtained from the performance evaluation show that the algorithm shows a wide advantage over the well-known Variable Step Size (VSS) in terms of convergence speed and error reduction.