Low-complexity Fast NLMS based on Partial Updating Strategy for Acoustic Echo Cancellation

For the purpose of reducing the computational load of the Fast Normalized Least Mean Square (FNLMS) algorithm for Acoustic Echo Cancellation (AEC) applications, we propose to incorporate the Partial Update strategy into the Fast NLMS algorithm to obtain a new algorithm that has less computational complexity and fast convergence rate than the original FNLMS algorithm. The proposed algorithm is denoted Partial Update-FNLMS (PU-FNLMS) algorithm. Experiments are carried out using stationary and non-stationary inputs filtered by an acoustic car impulse response. Simulation results demonstrated the good performances of the proposed algorithm compared to the NLMS algorithm, and this proposed algorithm has the same performance in the transient and permanent phase as the Full-Update FNLMS algorithm but with a lower arithmetic cost.