A double-talk detector based on generalized mutual information for stereophonic acoustic echo cancellation systems with nonlinearity

This paper aims at double-talk detector design for acoustic echo cancellers (AECs) enabling transmission of speech signals in stereo. A reliable double-talk detector (DTD) is often a necessary requirement for AECs. Its role is to freeze the adaptation of the AEC filter when near-end speech is present in order to avoid divergence of the adaptive algorithm. The existing DTDs based on the normalized cross-correlation vectors can not perform well when the system components such as loudspeakers exhibit nonlinearities. In this paper, we propose a novel DTD based on the generalized mutual information (GMI) and show that it performs well in the presence of nonlinearity. Effectiveness of the proposed method is demonstrated through computer simulations.