Non-linear echo cancellation - a Bayesian approach

Echo cancellation literature is reviewed, then a Bayesian model is introduced and it is shown how how it can be used to model and fit non-linear channels. An algorithm for cancellation of echo over a nonlinear channel is developed and tested. It is shown that this nonlinear algorithm converges for both linear and nonlinear channels and is superior to linear echo cancellation for cancelling an echo through a nonlinear echo-path channel.