An adaptive acoustic echo cancellation without double-talk detection

The paper proposes a novel adaptive echo cancellation (AEC) without double talk detection (DTD) that can work stably under all operating models. This breaks through the restriction that all of the adaptive echo cancellations must have DTD currently. The novel AEC is accomplished by injecting a low-level broadband random training signal to the far-end signal before transmission into the local room, modeling adaptively the echo path by taking the training signal as the reference signal, removing the speech signal interference of the adaptive algorithm by an adaptive prediction filter (APF), and modifying the distortion of the reference signal due to the APF by a compensation filter (CPF). The computer simulations confirm the AEC without DTD can provide excellent echo cancellation effect under all operating models.