Scalable perceptual audio coding with a hybrid adaptive sinusoidal signal model

This thesis is concerned with the development and optimization of a signal model for scalable perceptual audio coding. A hybrid, signal-adaptive model for audio consisting of sines + transients + noise (STN) is presented. The partial noise loudness (PNL) perceptual metric is newly investigated as a tool for control of model adaptation, and particularly for control of adaptive time segmentation in response to transient events. A perceptually-tuned algorithm is proposed for selection of sinusoidal STN model components. The proposed methodology minimizes the excitation pattern differences between the original and modeled signals on a short-time basis. The technique, known as Excitation Similarity Weighting (ESW), is shown to outperform existing methods for the selection of sinusoidal STN components, particularly for the sparse representations of the model parameters that are ultimately associated with low-rate coding applications. The STN model behavior under maximum signal-to-mask ratio (SMR) selection criterion is used to provide a benchmark against which ESW performance is compared. Preliminary performance evaluation results are given to facilitate the application of the STN model to scalable, high-fidelity audio coding.