Audio Dequantization Using (Co)Sparse (Non)Convex Methods

TL;DR

This paper explores audio dequantization techniques using sparse and non-convex methods, evaluating their effectiveness with SDR and perceptual metrics, and introduces new formulations in both synthesis and analysis frameworks.

Contribution

It introduces novel sparse and non-convex methods for audio dequantization, covering both synthesis and analysis variants, and provides comprehensive evaluation results.

Findings

Improved SDR scores over existing methods

Enhanced perceptual quality as measured by PEMO-Q

Validation of the effectiveness of non-convex approaches

Abstract

The paper deals with the hitherto neglected topic of audio dequantization. It reviews the state-of-the-art sparsity-based approaches and proposes several new methods. Convex as well as non-convex approaches are included, and all the presented formulations come in both the synthesis and analysis variants. In the experiments the methods are evaluated using the signal-to-distortion ratio (SDR) and PEMO-Q, a perceptually motivated metric.