A subband based acoustic echo canceler for a hands-free telephone

A novel subband-based acoustic echo canceller (AEC) for a handset-free cellular phone is proposed. The proposed scheme consists of critically down-sampled 16 DFT filter bank nonoverlapping main subbands and 32 DFT filter bank auxiliary subbands. The auxiliary subbands provide spectral coverage in the nonoverlapped regions. Four different adaptation schemes are considered: (i) the NLMS algorithm; (ii) predictive NLMS (PNLMS) algorithm; (iii) gradient lattice algorithm; and (iv) the thin gradient lattice algorithm. The performance of these algorithms are compared both by simulation and real-time experimentation. The results provided show that the proposed subband AEC with a thin lattice algorithm exhibits the fastest convergence speed as well as the highest reduction in the acoustic echo level in comparison to the NLMS, PNLMS, and the conventional gradient lattice algorithms.