A Variable Step-Size LMS Adaptive Filtering Algorithm

A varariable setp-size LMS algorithm which is based on the nonlinear functional relationship between the step-size and the error signal is presented. The new algorithm has better convergence properties at the beginning of adaptiving while ensuring small final misadjustment. Moreover the performance of the algorithm adapts to jump systems by using a hopparameter α(n) to adjust the step-size and eliminates the disturbances of independent noise adjusting the step-size according to the square of the timed-averaged estimate of the atuocorrelation of the current error singal e(n) and the previous error singal e(n-1). Computer simulation results confirm the algorithm is superior to the former algorithms in performance.