A new variable step size algorithm applied in LMS adaptive signal processing

This paper presents a kind of new variable step size algorithm in which the changing of step size is associated with error signal. The error signal is bigger during the initial working period of system, and step size is bigger too in order to satisfy the need for faster convergence of adaptive system, then the error signal and the step size is smaller too, so the system is always in a stable state of convergence. By adjusting the very only regulatory factor in the algorithm, the step size can be adjusted easily to meet the needs of different filtering. MATLAB results show that the convergence, stability and filtering effect of adaptive system when using this algorithm are superior to fixed step size.