An echo path variation detector based on coherence

A low cost method for detecting echo path variation is proposed. A detection parameter is constructed by averaging the coherence function between the far-end signal and the error signal over time and frequency. The detection delay caused by the time averaging is accounted for by setting different detection thresholds for the rising and dropping edges of the detection parameter. Numerical examples show that the proposed method can reliably differentiate between echo path variation and double talk.