The Fast-NLMS Algorithm for Acoustic Echo Cancellation with Double-Talk Detection

In this paper, we introduce a Fast Normalized Least Mean square algorithm (FNLMS) to eliminate the acoustic echo in the presence of double talk signal. When the near-end speech is present the performance of the adaptive filtering is degraded, thus, a double-talk detector (DTD) must be used for controlling the Acoustic Echo Cancellation (AEC). A Normalized cross correlation (NCC) method applied to FNLMS algorithm is proposed in this paper. Experimental results using different input signals and real measured impulse response (audio-conference room), show a fast convergence and detection of double-talk of the proposed algorithm compared with NLMS algorithm.