Abstract
The affine projection algorithm (APA) is widely used for acoustic echo cancellation, but faces challenges like slow convergence, high steady-state errors and sensitivity to impulsive noise. In this paper, a parametric Welsch APA has been introduced and integrated into the multiple sub-filter-based framework to address these issues for highly correlated input signals and impulsive noise. The employed parametric Welsch cost function robustly suppresses impulsive noise without any matrix inversion operation as in APA. On the other hand, a multiple sub-filter framework improves convergence and tracking performance. Through a theoretical stability and convergence analysis, the performance limits have been validated. Simulations performed using speech and synthetic signals with Bernoulli–Gaussian and impulsive noise demonstrate that the proposed approach achieves superior steady-state and convergence performance than existing APA variants.
Citation
Gagandeep Singh, Asutosh Kar, Vasundhara, Monorama Swain, Jesper Rindom Jensen, Mads Græsbøll Christensen
Robust Sub-Filter Based APA With Parametric Welsch Function for Acoustic Echo Cancellation Under Impulsive Noise
Electronics Letters, 62(1):e70595, doi:10.1049/ell2.70595, 2026.
BibTeX
@article{GagandeepSingh2026aco_echo,
title = {Robust Sub-Filter Based APA With Parametric Welsch Function for Acoustic Echo Cancellation Under Impulsive Noise},
author = {Gagandeep Singh and Asutosh Kar and Vasundhara and Swain, Monorama and Jesper Rindom Jensen and Mads Græsbøll Christensen},
journal = {Electronics Letters},
publisher = {Association for Computing Machinery},
address = {New York, NY, USA},
volume = {62},
number = {1},
pages = {e70595},
doi = {10.1049/ell2.70595},
url = {https://ietresearch.onlinelibrary.wiley.com/doi/abs/10.1049/ell2.70595},
year = {2026}
}