A new doubletalk and channel change detection algorithm based on hypothesis testing

In acoustic and network echo cancellation, the detection of doubletalk and echo path channel changes are important to the control of the echo canceller's adaptive filter. This paper investigates joint doubletalk and channel change detection from an M-hypotheses test perspective. Also, using a stationary Gaussian stochastic input signal model, we propose a doubletalk versus channel change detection algorithm based on the likelihood ratio test. This proposed detection algorithm intuitively has a dynamic threshold which is based on the probabilities of past doubletalk and echo path change detection outputs. Simulation results prove the efficiency of the proposed detection algorithm.