Detection of echo generated in mobile phones using pitch distance

This paper presents a method for detecting the presence of echo returned from a mobile phone and estimating its delay. The detector bases its decisions on the distance between pitch periods in uplink and downlink speech signals. We argue that the pitch estimation errors in AMR speech codec are well modelled by Laplacian distribution. Using a mixture of Laplacian and uniform model for errors, we then derive an algorithm to detect the presence of the echo and estimate its delay. Performance of the derived algorithm is investigated by simulations.