Techniques for Improving the Accuracy of Sinusoidal Tracking

Abstract

This paper proposes three novel techniques for improving the accuracy of the heuristic sinusoidal tracking algorithm proposed in [1]. When applied to audio coding these techniques extend the traditionally speech coding approach into true wideband audio coding. These techniques provide proper multiresolution sinusoidal tracking, matching across a number of variables instead of frequency alone, and global optimization of these variables across sinusoidal tracks. When these techniques are used together a heuristic tracking algorithm is created which has many of the benefits of tracking algorithms formulated as a Hidden Markov Model (HMM) problem, but at a far reduced computational and implementation complexity.