Gradient variable-step LMS self-adaptation filtering method

Abstract

The invention provides a gradient variable-step LMS self-adaptation filtering method. The method includes the steps that (1) input signals X(n)+{x(n), x(n-1), ..., x(n-m+1)} are signal vectors formed by delays at different moments, wherein x(n) is a sampling value of a first-order filter at the n moment, and m is the order of a transversal filter; (2) the input signals are multiplied by corresponding weighted values and then summation is conducted to obtain an actual system output value y(n), and all weight vectors are initialized to be zero; (3) y(n) is subtracted from d(n) to obtain error signals e(n); (4) smooth gradient vectors g(n) are obtained; (5) the smooth gradient vectors at the adjacent moments are multiplied to obtain iterative step parameters at the n moment; (6) a weighted vector at the moment is obtained; (7) cyclic computing is conducted from the first step to the sixth step, then iterative computation is conducted and results are output. Through the gradient variable-step LMS self-adaptation filtering method, fast convergence on the premise of not reducing steady state errors is achieved.

Description

technical field [0001] The invention belongs to the technical field of adaptive signal processing, and relates to an LMS adaptive filtering method with gradient variable step size. Background technique [0002] As a branch of digital signal processing, adaptive filtering technology has been widely used in real life. Among them, the Least Mean Square (LMS) algorithm has become one of the most concerned algorithms in the practical use of adaptive filtering because of its simple implementation and robustness to signal statistical characteristics. However, the main shortcoming of the classic LMS algorithm is the contradiction between the convergence speed and the steady-state error, which seriously affects its application in some systems that require high convergence speed. The classic LMS algorithm uses a fixed step size, and the step size parameter μ controls the robustness, convergence speed and steady-state error of the algorithm. Generally, if the step size parameter is l...

AI Technical Summary

Problems solved by technology

[0004] The purpose of the present invention is to provide a gradient variable step size LMS adaptive filtering method, which solves the problem that the steady-state error of the LMS algorithm and the improvement of the convergence speed of the algorithm are difficult to balance in the prior art

Images