An algorithm for discrimination and detection of the two phenomena double talk and abrupt changes in the echo path is proposed for fading channels. Being able to discriminate and detect these two phenomena is crucial since the echo canceler must react differently. The suggested detection scheme is based on a sequential detection approach. The communication channel is modeled as a randomly time-varying linear system. An autoregressive model is used to describe the time evolution of the channel taps. The channel parameters are identified using a Kalman filter coupled with a recursive least squares algorithm, and, based on model assumptions, the maximum a posteriori probabilities corresponding to double talk and abrupt echo path changes are calculated. The proposed scheme is verified experimentally by way of computer simulations.