A Variable Step Size Adaptive Filtering Algorithm with Feedback Mechanism

Aiming at the problem that the traditional fixed-step LMS (Least Mean Square) algorithm can not satisfy both faster convergence speed and smaller steady-state error, an improved variable-step LMS algorithm is proposed in this paper. In this paper, an improved variable step size LMS algorithm based on Sigimoid function is proposed.In this paper, the error signal of the previous time is introduced as a reference to control the shape and scope of the step factor curve. Compared with the error value of the current time, the feedback mechanism can make the system more stable. The simulation results show that, compared with other adaptive filtering algorithms proposed in recent years, the algorithm achieves steady state after 200 iterations on average, converges faster and runs faster by 7.2%. When the system mutates, it can quickly restore steady state again.