A combined LMS algorithm based on sliding variance decision and its performance analysis

Abstract

Based on the channel equalization model, the classic LMS algorithm and traditional variable step size LMS algorithms were discussed, and the deficiency that they cannot simultaneously establish optimal convergence speed and steady-state errors was pointed out subsequently. On this basis, a combined LMS algorithm based on sliding variance decision was introduced. In this algorithm, different step size strategies were adopted in different convergence stages to achieve better overall convergence performance through independent target optimization. A time-varying multipath channel model was constructed and the ability of the combined algorithm to suppress multipath interference in a practical communication system was simulated and analyzed. Simulation results demonstrate that the combined LMS algorithm can simultaneously realize the optimization of convergence speed and steady-state errors and has better overall convergence performance, which significantly improves the ability of practical communication systems to suppress multipath interference.