Investigation of a Simplified Approach for the Removal of Nonlinearly Distorted Acoustic Echos
|Supervisor:||Dr.-Ing. Marcus Zeller, Room 6.11c|
|Faculty:||Prof. Dr.-Ing. Walter Kellermann|
Acoustic echo cancellation has been an active field of research for many years now. Since one of the major challenges for the compensation mechanism is the ability to follow the rapidly time-varying impulse response of the acoustic enclosure, these devices are usually realized by adaptive linear filters.
However, due to the ongoing miniaturization of electro-acoustic hardware components (as e.g. in mobile phones), the behaviour of these devices increasingly exhibits also nonlinear input/output relations. These nonlinear distortions that are e.g. given by simple saturation effects, reduce the performance of purely linear adaptive filters significantly. For an effective compensation, it is therefore necessary to apply also nonlinear filter models for the adaptive echo canceller.
The task of this thesis is to implement and thoroughly evaluate a newly proposed method for removing the nonlinearly distorted residual echos of a hands-free communication scenario. The setup of this concept is based on the combination of a purely linear adaptive filter and a subsequent correlation model of nonlinear signal components in the frequency domain. A special focus of this work is given by the question if this approach can be used to estimate and remove the nonlinear residual echo caused by different types of nonlinearity (hard limiter, saturation characteristics, Volterra filters, measured data). Moreover, investigations on possible extensions of this approach lie also within the scope of this work.