Robust nonlinear acoustic echo cancellation using a metaheuristic optimization approach

Nonlinearities in audio systems are caused by different nonlinear components like amplifiers and speakers. The nature of these impairments makes acoustic echo cancellation (AEC) harder to achieve, requiring the use of advanced and complex algorithms to offer satisfying performance. Adaptive filters in combination with nonlinear mapping schemes like Volterra and Hammerstein are widely used for this purpose. In this work, we address the AEC problem in the presence of severe nonlinear distortion. The proposed method based on metaheuristic optimization uses a genetic algorithm (GA) to estimate the nonlinear function parameters and the room impulse response. The method is compared to a reference technique that uses a sigmoid transform approach in conjunction with the recursive least square (RLS) algorithm. Simulation results show high robustness of the proposed approach to strong nonlinearities and saturation effects compared to the reference method.