A mutual information based double-talk detector for nonlinear systems

A reliable double-talk detector (DTD) is often a necessary requirement for acoustic echo cancellers (AECs). Various techniques have been proposed in the literature to detect double-talk for AEC systems. In those DTD algorithms, the acoustic echo path is assumed to be linear. On the other hand, nonlinear AEC is necessary when the system components such as loudspeakers exhibit nonlinearities. In this paper, we show that existing DTDs do not perform well when the acoustic echo path is nonlinear. We propose a novel DTD based on the mutual information (MI) concept and show that it performs well in the presence of nonlinearity. Effectiveness of the proposed method is demonstrated through computer simulations.