Franck Beaucoup

Nonlinear System Identification Using a Subband Adaptive Volterra Filter

This paper proposes a flexible and efficient subband adaptive second order Volterra filter structure for nonlinear system identification. Acoustic echo cancellation is an application of system identification that is critical in hands-free telephony, for which a linear model is usually assumed. However, echo cancellation is limited by inherent system nonlinearities, of which loudspeaker distortion is one of the main sources. Experimental results, obtained in an office room environment, show that the proposed structure was able to model the nonlinear room transfer function more accurately than a linear only model by up to 10 dB lower mean square error, at a comparable computational complexity.

Minimum-distortion-based self-steering algorithm for microphone arrays using switched fixed beamforming

This article presents a new self-steering algorithm for microphone arrays in audio communications devices such as desktop phones, mobile handsets and multimedia tablets. The general framework is that of fixed beamforming on a predefined set of look directions to cover the spatial range of interest, followed by post filtering. The proposed algorithm improves on existing algorithms by introducing a feedback mechanism between signal-path beamforming and talker localization instead of the conventional two-step approach, with the latter implemented as an independent preprocessing step to the former. The new algorithm is shown to be an extension of the popular Steered Response Power (SRP) algorithm and is compatible with traditional decorrelation techniques such as Phase Transform (PHAT).

Delayed adaptation for improved doubletalk resilience in adaptive echo cancellers

This paper proposes a simple modification to adaptive echo cancellers to prevent rapid divergence due to doubletalk conditions. The proposed structure introduces a delay only into the adaptation algorithm, which compensates for doubletalk detector response time at the onset of near-end speech. The structure is described and analyzed for the NLMS and cross-correlation-based doubletalk detector algorithms. The method introduces no delay into the return path, and no additional constraints on adaptation step size arise. Simulations with ITU-T Q. 168 tests confirm that the proposed structure prevents divergence at the onset of doubletalk for near-end to echo signal ratios of-26 to -6 dB.