Stereo echo cancellation algorithm using imaginary input-output relationships

A stereo teleconferencing system provides greater presence in teleconferencing compared with a monaural system. It helps listeners distinguish who is talking at the other end by means of spatial information. The most important problem in stereo echo cancellation is that the adaptive filter often misconverges or, if not, its convergence speed is very slow because of the crosscorrelation between stereo signals. A new stereo echo cancellation algorithm using imaginary input-output relationships is proposed. This algorithm is based on the idea of approximating the input-output relationships for the reversed stereo signals. Unlike conventional methods, which use the input-output relationships in only one situation, our algorithm uses the simultaneous equations for the input-output relationships under two different conditions. Consequently, convergence to the unique steady-state solution is faster than with the conventional algorithm. Computer simulations using data recorded in a conference room demonstrate the effectiveness of this algorithm.