Fast convergence LMS adaptive filters employing fuzzy partial updates

This paper presents a method to improve the performance of reduced calculation adaptive filters. We use the sequential partial update method to achieve low computation complexity. Furthermore, we include the variable step-size approach to aim last convergence. The variable step size approach is based on a fuzzy method to determine the appropriate step-size on each iteration step. By using the proposed method, the adaptive filter converges faster while pretending the steady state error as the previously proposed reduced calculation adaptive filler. The instantaneous step size is determined from the present square of the error signal to produce sudden changing. Additional rule or conditions are included to prevent the adaptive algorithm to become unstable. Simulation results are presented to compare the performance of the new approach, the fixed step-size LMS algorithm and sequential partial update LMS (S-LMS) algorithms.