System Identification Problems on Acoustic Echo Cancellation

- Nonlinear Extensions for Practical Problems -

Abstract

System identification theory is now well established. However, its framework may still be modified or extended depending on actual applications. Acoustic echo cancelation is a reasonable field of application for system identification. In this paper, some recent results of extending the framework for acoustic echo cancelation are reviewed, in which nonlinearity is a keyword. To improve robustness during double-talk or noisy situations, some nonlinearly modified error cost functions have been introduced. To cancel distorted echo, nonlinear echo path modeling and its identification algorithms have been developed.