Adaptive affine combination of two IPNLMS filters for robust sparse echo cancellation

Conventional adaptive filtering algorithms (lms,nlms) are widely used in system identification applications. however,the performance of these algorithms is degraded when the system echo path is sparse in nature as in network and acoustic echo cancellation scenarios. proportionate-type filters (ipnlms) are considered as suitable candidate to achieve better performance for sparse echo paths,but they fail to exploit the time varying system sparsity. moreover,there exists some compromise between their convergence speed and steady-state error. to overcome these limitations,a combination approach of two adaptive filters that combines the output of individual filters through a mixing parameter has been developed. we propose an adaptive affine combination of two ipnlms filters as a robust solution to alleviate the convergence speed vs steady-state error tradeoff,and to efficiently increase ipnlms robustness to time varying sparsity of the system. effectiveness of our proposed affine combination approach has been validated from the matlab simulations.