A Family of Adaptive Filtering Algorithms for Monophonic and Stereophonic Acoustic Echo Cancellation

Abstract

In hands-free devices, acoustic echo arises due to the coupling between the microphone and loudspeaker. The acoustic echo cancellation (AEC) algorithms proposed in the literature assume that the echo path is linear. However, the transducers like loudspeakers, microphones used in acoustic devices induce the nonlinear distortion into the echo path. This distortion affects the convergence, the trade-off between convergence and steady-state, echo return loss enhancement performance of the AEC. To address the issue of nonlinear distortion in AEC, the nonlinear acoustic echo cancellation (NAEC) algorithms based on Volterra and polynomial have been proposed in the existing literature. There is a scope for improving the convergence and trade-off between steady-state and convergence for these approaches. On similar lines, the functional link adaptive filtering (FLAF) model-based NAEC algorithm has been proposed, which falls under the linear-in-parameter category and depends on trigonometric functions for nonlinear distortion modeling. This approach brings substantial improvement over the existing Volterra and polynomial-based models in terms of overall nonlinear acoustic echo cancellation. But, there exists an opportunity to address the slow convergence and further improvements in steady-state without degrading the echo return loss enhancement.