Adaptive filtering method

Abstract

The present invention provides an adaptive filtering method which comprises the steps of (1) carrying out momentum LMS iterative filtering, and using a gradient descent method to adjust a weight vector, (2) carrying out variable momentum factor LMS iterative filtering, before iteration reaches convergence or in a non-stationary environment, through adaptive adjustment of a combination factor such that a step length is relatively large, thus improving filtering method convergence speed, and allowing the step length to be relatively small and reducing a steady-state error to improve stability when reaching a convergence state, (3) selecting different step length variable momentum factor systems, (4) carrying out double variable momentum factor filtering, (5) carrying out combination factor iteration, and (6) obtaining a change result, realizing filtering, obtaining an expression method of a mean square error and the iteration modes of the momentum factor and the combination factor, and thus obtaining the convergence characteristic of the mean square error and the adaptive change result of the momentum factor and the combination factor.

Description

technical field [0001] The invention belongs to the field of adaptive filtering, and in particular relates to improving the convergence speed of the adaptive filter and reducing the steady-state error of the filtering method, and a filtering method capable of balancing and optimizing the performance of the two aspects. Background technique [0002] Adaptive filters play an important role in communication fields such as echo cancellation, automatic equalization, and radar sonar beamforming, as well as in other signal processing fields such as identification parameters, noise suppression, and spectral estimation. In these fields, in practical application problems, the received signal acquired by the receiving device is often accompanied by interference and noise caused by the environment, and the resulting increase in the signal bit error rate will significantly affect the accuracy of the acquired signal. Interference and noise exist in almost all application fields in reality...

AI Technical Summary

Problems solved by technology

[0003] In the previously proposed adaptive filtering method for the variation factor, the convergence speed and steady-state error of the LMS method are controlled by changing the momentum factor, in order to improve the convergence speed while keeping the steady-state error small. However, due to the limitation of the fixed step size parameter, the final steady-state error is ideally about equal to the steady-state error of the traditional LMS method, so we say that its effect still has a certain bottleneck, and the filtering method cannot be larger Minimized steady-state error reduction

Images