Subband adaptive filter trained by differential evolution for channel estimation

TL;DR

This paper introduces a differential evolution-based variant of the normalized subband adaptive filter (NSAF) to improve convergence speed in channel estimation, addressing limitations of traditional NSAF.

Contribution

The paper proposes a novel DE-NSAF algorithm that enhances convergence rate for adaptive filtering in channel estimation, considering the fullband adaptive filter model.

Findings

DE-NSAF converges faster than traditional NSAF.

Simulation results show improved performance in channel estimation.

The algorithm effectively searches for the global optimal weight vector.

Abstract

The normalized subband adaptive filter (NSAF) is widely accepted as a preeminent adaptive filtering algorithm because of its efficiency under the colored excitation. However, the convergence rate of NSAF is slow. To address this drawback, in this paper, a variant of the NSAF, called the differential evolution (DE)-NSAF (DE-NSAF), is proposed for channel estimation based on DE strategy. It is worth noticing that there are several papers concerning designing DE strategies for adaptive filter. But their signal models are still the single adaptive filter model rather than the fullband adaptive filter model considered in this paper. Thus, the problem considered in our work is quite different from those. The proposed DE-NSAF algorithm is based on real-valued manipulations and has fast convergence rate for searching the global solution of optimized weight vector. Moreover, a design step of new algorithm is given in detail. Simulation results demonstrate the improved performance of the proposed DE-NSAF algorithm in terms of the convergence rate.