A fast method for the NLMS algorithm with time‐variant decorrelation filters

Abstract For many applications, such as the cancellation of room echoes, the NLMS algorithm converges too slowly because of the correlatedness of the adjustment signals. Therefore, it was repeatedly proposed to increase the convergence speed by means of decorrelation filters. In the present contribution a well‐known structure based on the use of time‐variant decorrelation filters is replaced by an alternative one which supplies exactly the same results but allows the computational efforts to be reduced considerably.