Sparseness-Controlled Affine Projection Algorithm for Echo Cancelation

Affine projection algorithms (APAs) are popular in echo cancellation due to its fast convergence. Proportionate APA (PAPA) and improved APAP (IPAPA) have been proposed recently to enhance the performance of APAs based on the concept of assigning different effective step-sizes to the tap coefficients. Conventional IPAPA is found to be able to outperform PAPA by using a proper control parameter. In this paper, two sparseness controlled IPAPAs (SC-IPAPAs) are proposed to further improve the convergence of IPAPA, namely, SC-IPAPA-I and SC-IPAPA-II. Unlike conventional IPAPA, the control parameter is a function of the sparseness measure of the estimated impulse response in the proposed SC-IPAPA-I and SCIPAPA-II, rather than a predefined constant. The contribution of our proposed SC-IPAPA-II is to remove the need to decide on a predefined control parameter as well as achieving high rate of convergence that is robust to the sparseness of the impulse response. Simulation results show that SC-IPAPA-I and SCIPAPA-II have higher rate of convergence than that of IPAPA.