Description
Date depot: 30 août 2019
Titre: Drum sound synthesis with generative adversarial neural networks
Directeur de thèse:
Axel ROEBEL (STMS)
Domaine scientifique: Sciences et technologies de l'information et de la communication
Thématique CNRS : Non defini
Resumé:
The project proposes to investigate into drum synthesis with GAN. A key hypothesis of the present
project is that GAN style audio synthesis has the potential to unlock creative sound effects that can
significantly enlarge the reservoir of sounds playable with electronic drums. For this to take place
audio synthesis with GAN synthesizers needs to make significant progress. The present project aims
to develop methods that achieve sound quality and variability that is at a similar level than what has
been achieved for image synthesis. The following questions have to be solved:
• An appropriate sound representation (spectral or temporal domain) for synthesis of drum
sounds;
• a synthesis method allowing for professional quality audio with at least a sample-rate of
44.1kHz;
• control structures that are establishment of control interfaces.
Doctorant.e: Lavault Antoine