Closed-loop tracking of sinusoids for speech and audio coding

A well recognised problem in low bit rate representation of audio and speech signals, based on the sinusoidal model, is that of tracking the sinusoidal components. Imperfections in the analysis process and the presence of components over a limited duration of time gives rise to ambiguities in the tracking process. As a solution to this problem, we propose a mechanism to achieve closed-loop tracking by means of using analysis-by-synthesis incorporating phase prediction. A simple implementation of such an algorithm is discussed by considering an overlap-add synthesizer. Finally, the results are presented using a voiced speech segment as an example.