Improving Acoustic Echo Cancellation by Exploiting Prior Knowledge of RIR Energy Decay

Pedro de Carvalho Cayres Pinto, Roberto Campos, Mariane R Petraglia

DOI: 10.14209/sbrt.2024.1571036912

Evento: XLII Simpósio Brasileiro de Telecomunicações e Processamento de Sinais (SBrT2024)

Keywords: Adaptive filters Acoustic echo cancellation Normalized least-mean-squares algorithm Affine Projection Algorithm

Abstract

Acoustic echo cancellation (AEC) through adaptive filters requires a large number of coefficients to establish good communication in reverberant environments. In order to increase the convergence rate of the adaptive algorithm, in this paper we explore the exponential decay rate of the room impulse response (RIR) by introducing an additional constraint, based on the energy relationship between blocks of coefficients, to the optimization problem. The performances of the proposed algorithms are investigated in different scenarios and compared with the performance of traditional algorithms employed in the AEC application. We also examine the influence of the algorithms' parameters in the convergence rate and final solution.

Download