Dynamic tap-length estimation based low complexity acoustic echo canceller

A low complexity Multiple Sub-Filters (MSF) parallel structure for monophonic acoustic echo cancellation (AEC) has been proposed where order of each MSF has been estimated dynamically using a new variable step least mean square (VLMS) based fractional order selection algorithm. The proposed length selection algorithm improves the overall performance of the adaptive filter searching the optimum order of each MSF dynamically with faster convergence in a time varying environment. The dynamic order estimation algorithm adapt the MSF to the desired optimum length makes the variable order adaptive filter more efficient by reducing the overall complexity in design and the adaptation noise. The performance analysis depicts that the proposed algorithm converge to the optimum value in mean. The convergence performance of the MSF based parallel structure is studied for common error and different error adaptive algorithms. Simulation results show that MSF with both adaptation algorithms provides better convergence and tracking performance over the conventional echo canceller i.e. a single long filter (SLF). The order adaptation reduces the total number of weights to model the system hence reduce the complexity in design and avoids the adaptation noise.