Online Steepest Descent Optimization of Muting Technique Parameters in ITU-T G.722 Frame Erasure Concealment

In this paper, we propose an enhanced adaptive muting algorithm using a logistic function based on the steepest descent optimization criterion for frame erasure concealment in ITU-T G.722. Since our previous approach uses a sigmoid function as the muting curve and core parameters are determined globally according to the overall training data in the separate training step, i.e., the grid search, temporal variations cannot be considered over time. To address this problem, we propose to update the core parameters of the logistic function instantaneously using the least mean square (LMS) algorithm during good frames and to use the parameters in the missing frames. To perform the LMS algorithm on each frame, the steepest descent optimization criterion for the logistic function is applied to minimize the squared error between the original signal and the reconstructed signal during good frames. Experimental results indicate that the proposed muting approach outperforms the conventional muting approaches.