Convex Combination Of Adaptive Linear Finite Impulse Response And Nonlinear Volterra Filters For Acoustic Echo Cancellation

We all know how much joy and freedom hands-free communication system has brought to ourrndaily life. But associated with this is the problem of acoustic echo which degrades the quality ofrncommunication. A number of researches have been conducted in order to eliminate acousticrnecho. Most of the acoustic echo cancellers are made on the assumptions of the linearity of thernecho. However, this assumption is no longer valid when the nonlinearity along the echo path isrntaken into account. This led to the invention of Nonlinear Acoustic Echo Cancellation (NLAEC)rntechniques such as Volterra filters. However, in some situations, NLAEC is inferior to a linearrnAcoustic Echo Cancellation (AEC), especially when the channel generates a negligible (or no)rnnonlinear echo. In general, the ratio of the linear to nonlinear echo signal power is unknown arnpriori, and will vary over time, thus making it difficult to know if NLAEC would improve orrndegrade the cancellation. Therefore, in this thesis a convex combination of the linear and thernnonlinear Volterra filters is studied. Simulation has been carried out and the results indicate therncombination scheme follows the best contributing filter regardless of the nature of the echo.rnAnother important problem of AEC is the problem of double talk (DT) which occurs when bothrnparties talk at the same time. If DT is not controlled it results in diverging of the adaptive filterrncoefficients and as a result audible echoes pass without being cancelled. In order to alleviaternthis problem most DT controllers have been studied based on correlation of the availablernsignals. Since correlation is valid only for linear relationship; this assumption doesn’t hold whenrnthe nonlinearity along the echo path is considered. Therefore, in this thesis mutual informationrn(MI) based DT detection has been studied together with AEC. Simulation is carried out byrnintroducing a DT region and the results obtained indicate performance improvement of 97.06%rnin terms of algorithm misalignment error (AME).rnKey words: Acoustic echo, In general, Convex Combination, Volterra, adaptive filters

Get Full Work

Report copyright infringement or plagiarism