The AQUA-RIUS project will last 42 months, starting January 1, 2023. Its objective is the study automatic analysis and processing of audio quality in audio recordings. Audio quality is defined in the project as the set of sound characteristics of a musical recording that are not dependent on the sound sources. This means that audio quality depends on mixing effects (such as filtering, dynamics compression, reverberation) and degradation (e.g. addition of noise, lossy audio coding, saturation).

This project intends to analyze and describe the audio quality and degradations (from a signal and psycho-acoustic point of view), and to simulate and modify the audio quality, including for example the remixing of a recording. Taking advantage of recent advances in machine learning, deep learning will be fully exploited, including the development of a differentiable library of audio processing, and the learning of generative networks.

The project partners are: IBISC/University of Evry (Dominique Fourer the project coordinator, Hichem Maaref), IRCAM (Rémi Mignot, and Diemo Schwarz), and Télécom-Paris (Geoffroy Peeters).