Behavior of variable step size NLMS algorithms for acoustic echo cancellation

The paper presents an acoustic echo cancellation application of three variable step-size Normalized Least Mean Square (NLMS) algorithms: the simple VSS-NLMS (SVSS-NLMS), the new non-parametric (NEW-NPVSS-NLMS) and the practical VSS-NLMS (PVSS-NLMS) algorithms. Simulations were performed in two scenarios: single-talk and double-talk. We will prove that the PVSS-NLMS algorithm can be used in a double-talk situation without using a double-talk detector (DTD). The capabilities and performances of the algorithms are analyzed in terms of convergence rate, tracking and robustness, but also from other practical point of view, such as available parameters.