Performance analysis of an improved split functional link adaptive filtering algorithm for nonlinear AEC

In the process of nonlinear acoustic echo cancellation (NAEC), the eradication of echo becomes challenging due to the nonlinear distortion introduced by the low-cost hands-free communication devices. To mitigate the effect of these artifacts, various NAEC adaptive filtering algorithms are existing which find their applicability in different echoed environments. But, there exists a scope to improve the echo return loss enhancement (ERLE) performance and the rate of convergence for adaptive NAEC algorithms. These improvements can be achieved with further optimization of key parameters in existing mean square error based adaptive algorithms and by proposing a novel combination of linear and nonlinear adaptive filters. In this paper, a split functional link-based adaptive filter (SFLAF) is proposed with an improved optimized-normalized least mean square adaptive algorithm for NAEC. In addition to that, the convergence and steady-state analyses of the proposed algorithm were presented in this paper. The performance of the proposed algorithm is compared to existing SFLAF based NAEC algorithms, and improvements are presented. The colored noise signal, clean speech, and speech signal corrupted with white noise are subjected as inputs to the proposed NAEC framework under different signal-to-noise ratios. The ERLE, spectrograms, and perceptual evaluation of sound quality are used as the performance indices for the comparison of the proposed algorithm with its counterparts. An average improvement of 3 dB is observed in the case of the proposed algorithm compared to its counterparts. (C) 2020 Elsevier Ltd. All rights reserved.