Modeling External Volume Changes in Stereo Echo Cancellers

The paper describes a new algorithm for modeling and effectively compensating for the changes in external volume for a stereo acoustic echo canceller (AEC). The context is one in which the AEC connects to an external amplifier whose volume, tone, or Left-Right balance settings can be arbitrarily changed by the user. This change would normally result in the canceller losing convergence, and an echo to be heard temporarily at the far end. The proposed method provides a quick correction for the loss of convergence by applying two optimal complex gains to the output or to the coefficients of the L-R canceller filters, estimated according to an optimality criterion of minimizing the mean square error (MSE) at the combined canceller output. A two-path, subband architecture consisting of four cancellers per band is presented. The two complex gains are applied to the foreground canceller filters, while the background canceller continuously adapt using the two-channel extended least mean squares (LMS). Various conditions are used to ensure the validity of the computed gains, the uniqueness of the solution, and the overall system stability. Measurements of the echo return loss enhancement (ERLE) as well as perceptual evaluation of speech quality (PESQ) score show a significant improvement of the overall perceptual performance.