A Sparseness-Controlled Improved Proportionate Affine Projection Algorithm

The improved proportionate affine projection algorithm (IPAPA) can improve the convergence speed for the identification of sparse systems as compared to their conventional counterparts. However, its performance for non-sparse systems can be relatively poor, even slower than the APA algorithm. In this paper, a new algorithm is proposed to incorporate the sparseness into the IPAPA algorithm (SC-IPAPA) so that its performance for non-sparse impulse response is improved. The proposed approach first estimate how sparse is the target impulse response, then automatically adjust the related parameters to assure their convergence. Consequently, when the impulse response is sparse, the algorithm allocates more weight to the proportionate term. For comparatively less sparse impulse responses, the algorithm aims to achieve the convergence of APA by applying a higher weighting to the APA term. Simulations with real speech signals show that proposed algorithm (SC-IPAPA) is better than IPAPA and SC-IPNLMS for impulse responses with different sparseness.