Variable Step Size Least Mean p-Power Algorithm Based on Improved Softsign Function

Abstract

The conventional adaptive filtering algorithm in signal processing with fixed step size will lead to its stability and convergence unable to be combined at the moment. To deal with this problem, the least mean p-power (LMP) algorithm is improved, and a variable step size least mean p-power algorithm based on an improved softsign function is proposed. The algorithm uses the improved softsign function to construct the variable step size function, while the moving weighted average method is applied to update the step size and keep the efficiency of the algorithm stable. Simulation experiments indicate that the improved variable step size LMP algorithm further reduces the steady-state error of the algorithm while maintaining the original convergence speed, thus better balancing the stability and convergence of the algorithm under the interference of ocean pulse noise compared with the existing fixed step size and variable step size algorithms.