On computationally-efficient NLMS-based algorithms for echo cancellation applications

Abstract

This paper investigates computationally-efficient adaptive algorithms suitable for echo cancellation applications. Power-of-two (PO2) quantization for the error and/or regressor vector is added to a selective-partial-coefficient-update algorithm (SPCUA) to further reduce the amount of computations. A simple stop-and-go procedure is included to save more computations when the generated error magnitude is below a certain limit. Simulation results with comparisons are included to show the advantages of the improved algorithms