A Data Selective Segment Proportionate Affine Projection Algorithm for Echo Cancellation

Abstract

The segment proportionate normalized least mean-square algorithm (SPNLMS) has been proposed with the objective of improving the adaptation convergence rate when modeling high-order sparse impulse response systems as compared to their conventional counterparts.. When the excitation signal is colored, especially the speech, the convergence performance of SPNLMS algorithm demonstrates slow convergence speed. The segment proportionate affine projection (SPAP) algorithm is an useful adaptive filter to improve the convergence speed of PNLMS-type filter by updating the weight vector based on several previous input vectors, Unfortunately, the SPAP algorithm obtains a faster convergence of the SPNLMS comes at the expense of an increase in the computational complexity linked to the amount of reuses employed. For this reason, the idea of proportionate adaptation combined with the framework of set-membership filtering in an attempt to reduce computational complexity of algorithm is presented. The proposed algorithm allows the reduction of the frequency of updates of the filter coefficients, where the filter coefficients are updated such that the output estimation error is upper bounded by a pre-determined threshold. Simulations show that its overall complexity is lower compared to fully updated SPAP and its performance is close to that of the SPAP algorithm.