Embodiments of a multi-channel audio communication device and methods for reducing echoes are generally described herein. Other embodiments may be described and claimed. In some embodiments, echo-reduction circuitry may insert training signals into digital audio signals at or below a noise floor and in a non-audible portion of the frequency spectrum based on a spectral mask. Training signals may be generated from spectrum estimates and training sequences. An adaptive filter may generate echo-cancellation signals using filter coefficients generated fawn the training sequences and return-path signals. The echo cancellation signals may remove echo signals from the return-path signals.