An efficient quasi LMS/Newton adaptive algorithm for stereophonic acoustic echo cancellation

In this paper an efficient quasi LMS/Newton adaptive algorithm is proposed for stereophonic acoustic echo cancellation. This method employs an efficient pseudo-diagonalization approach to the estimated joint-input correlation matrix with an emphasis on reducing its high cross-correlation components. We derive an estimate of the inverse joint-input correlation matrix by means of this pseudo-diagonalization method. Simulation results show the improvement in convergence performance of this algorithm compared to the nonlinear RLS and nonlinear NLMS adaptive algorithms.