Please use this identifier to cite or link to this item:
Title: Nonlinear Acoustic Echo Cancellation Using Adaptive Algorithms under Noisy Environment
Authors: Sharma, Jashu
Supervisor: Kohli, Amit Kumar
Keywords: Acoustic-echo-cancellation (AEC);ERLE;RLS
Issue Date: 11-Aug-2017
Abstract: This research work presents a nonlinear-acoustic-echo-cancellation (NAEC) technique to tackle sigmoid-type nonlinearities under the noisy environment. The performance of NAEC technique is investigated for both cases, in the presence of environmental noise and in the absence of environmental noise. The nonlinear echo in acoustic systems is inevitable due to the inherent nonlinear characteristics of amplifiers and/or loudspeakers, which deteriorates the quality of speech as well as audio signal reception. Here, the sigmoid-type nonlinearity is modelled by using two control parameters, which determine the shape and clipping value of the saturation curve. These control parameters are adjusted by using variable-step-size (VSS) least-mean-square (LMS) adaptive algorithms to enhance the convergence rate and tracking capability, under the noisy environment. Two VSS-LMS algorithms are utilized, which are RVSS-LMS (as proposed by Kwong and Johnston; IEEE Trans. Sig. Proc., 1992) and TMVSS-LMS (as proposed by Aboulnasr and Mayyas; IEEE Trans. Sig. Proc., 1997) algorithms. The impulse response of acoustic echo path in a room is modelled as a tap-delay-line finite-impulse-response (FIR) filter, whose tap-coefficients are estimated by using the normalized-least-mean-square (NLMS) and recursive-least-square (RLS) algorithms at the different values of signal-to-noise-ratio (SNR), when the correlated as well as uncorrelated input signals are processed. The uncorrelated input signal samples are considered to be Gaussian random variable with zero-mean and unity variance. The correlated input signal samples (with zero-mean and unity variance) are assumed to follow a first-order autoregressive process i.e., AR(1) process in one case and a second-order autoregressive process i.e., AR(2) process in other case. In the research work presented in this thesis, only single talk case is considered in all input cases. Simulation results are presented to demonstrate the efficiency and efficacy of the NAEC technique using the adaptive algorithms in terms of the fast convergence rate and the high value of echo-return-loss-enhancement (ERLE) factor. The presented results connote that the NAEC technique based on TMVSS-LMS and RLS algorithm provides best results, in comparison to the RVSS-LMS and RLS algorithm based NAEC strategy. However, the fixed-step-size (FSS) LMS and NLMS algorithms fail to outperform the VSS-LMS and RLS algorithm based NAECs under similar conditions. It is evident from results that TMVSS-LMS and RLS algorithm based NAEC performs exceptionally well in case of the AR(2) correlated input, as compared to the AR(1) correlated and Gaussian uncorrelated inputs.
Description: Master of Engineering -ECE
Appears in Collections:Masters Theses@ECED

Files in This Item:
File Description SizeFormat 
4638.pdf1.4 MBAdobe PDFThumbnail

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.