Project Description
AQUA-RIUS ANR PRCE Project: “Audio Quality Analysis for Signal Representation, Indexing and Unification”
Scientific summary
Characterizing audio quality from a signal is an open problem of interest for the development of robust signal processing methods, but also in machine learning for the creation of trained models invariant to undesirable transformations. Thus, knowledge of these transformations enables the development of suitable signal enhancement methods (e.g. Wiener filtering, deconvolution, etc.), but can also be exploited to augment data to improve the performance of a machine learning system.
The AQUA-RIUS project postulates that audio quality can be defined as a combination of transformations in mathematical operators applied to the signal from its creation to its broadcast. These effects are not directly linked to the signal content and correspond, for example, to possible additive noise or artifacts linked to lossy compression (e.g., mp3, ogg).
Thus, this project has the following three objectives: 1) Objective characterization of audio quality with the development of techniques that detect individually or simultaneously all effects applied to a signal with their parameters 2) Simulation and synthesis of audio quality with its applications in data augmentation and domain adaptation, but also in artistic sound processing. 3) The cancellation and inversion of audio effects, with applications for restoring degraded audio signals.
The expected impact of this project is a better understanding of audio quality, better control of data augmentation practices for machine learning-based tools, and the emergence of new techniques for restoring audio signals and applying artistic effects to sound and music processing.
- Project start date: 01/12/2022
- Project duration: 42 months
- Project coordinator: Dominique FOURER (MCF Univ. Évry, IBISC team SIAM)
- Other IBISC members involved in the project: Hichem MAAREF (PR IUT Évry, IBISC team SIAM)
- IBISC institutional partners: Telecom Paris (LTCI), IRCAM (UMR STMS)
- Total budget: 510 K euros
- IBISC allocation: 149,16 K euros
- ANR project web page
Leave A Comment