Acoustic echo cancellation based on a recurrent neural network and a fast affine projection algorithm

To perform real time nonlinear adaptive filtering, we propose in this paper a structure based on a recurrent neural network in cascade with a fast affine projection (FAP) algorithm. The FAP algorithm allows a large set of tradeoffs between convergence rate, residual error, tracking capacity, and arithmetic complexity. Hence, the proposed structure has the potential for solving difficult nonlinear adaptive signal processing tasks; such as system identification where nonlinearity and nonstationarity are both important factors. The adaptive signal processing capability of the proposed structure has been tested in the context of the acoustic echo cancellation.