Neural Synthesis of Footsteps Sound Effects with Generative Adversarial Networks

Comunità, M., Phan, H., Reiss, J. D. - AES Convention - 152 (May 2022)

Abstract

Footsteps are among the most ubiquitous sound effects in multimedia applications. There is substantial research into understanding the acoustic features and developing synthesis models for footstep sound effects. In this paper, we present a first attempt at adopting neural synthesis for this task. We implemented two GAN-based architectures and compared the results with real recordings as well as six traditional sound synthesis methods. Our architectures reached realism scores as high as recorded samples, showing encouraging results for the task at hand.

Resources

• Paper: https://aes2.org/publications/elibrary-page/?id=21696
• Code: https://github.com/mcomunita/hifi-wavegan-footsteps
• Webpage: https://mcomunita.github.io/hifi-wavegan-footsteps_page
• Presentation: comunita2022hifiwavegan-presentation.pptx
• Poster: comunita2022hifiwavegan-poster.pdf

Citation

Comunità, M., Phan, H., Reiss, J. D. "Neural Synthesis of Footsteps Sound Effects with Generative Adversarial Networks" - Audio Engineering Society Convention 152. May, 2022.