Abstract

Addressing the challenge of sparse acoustic channel is important in communication systems such as tele- and video-conferencing systems. Sparse impulse response plays an efficient role in acoustic identification systems, particularly in long acoustic rooms. This article presents an enhanced version of the Improved Proportionate Normalized Least-Mean-Squares (IP-NLMS) algorithm. It adapts only the active coefficients, thereby reducing the computational complexity of IP-NLMS algorithm. The selected active coefficients of the IP-NLMS algorithm (SAC-IP-NLMS) are proposed exactly in sparse impulse response (SIR) in order to reduce the computational complexity with faster convergence rate. Several simulations conducted across various sparse environments - based on the time evolution of error signals, the mean square error, the echo return loss enhancement and the computational complexity - validate the effectiveness of the proposed algorithm.

Key words: Adaptive filtering; Sparse impulse responses; Improved Proportionate Normalized Least-Mean-Squares algorithm; Communication system.