A novel cost-effective sparsity-aware algorithm with Kalman-based gain for the identification of long acoustic impulse responses

Abstract

In this paper, a new robust sparsity (or sparseness)-aware adaptive filtering algorithm is proposed for the purpose of system identification and acoustic echo cancelation. It is named the improved proportionate fast normalized least mean square (IPFNLMS) algorithm. This latter has been derived by an effective integration of the update control matrix of the improved proportionate NLMS (IPNLMS) algorithm to the Kalman-based adaptation gain of the fast-NLMS (FNLMS) algorithm. Simulations were carried out both in synthetic and real long acoustic impulse responses at different sparseness levels with stationary and non-stationary inputs, followed by a verification with real experiment data. Results have shown interesting improvements for the proposed algorithm with respect to its ancestors in terms of convergence speed, steady-state performance, tracking capability and robustness against system sparsity variation.