Etienne Richan’s work on sound visualisation has been accepted at Timbre is a Many-Splendored Thing, which will take place from July 4th to 7th, 2018. There, he will present his software tool which aims to contribute to the study of audiovisual correspondences as well as assist the recognition of sounds with visual cues.
His research project uses an algorithmic approach for the extraction of perceptual features from musical sounds. This data is combined with the generative power of neural networks, which allow for the synthesis of a large variety of textural images. Neural networks as well as other machine learning algorithms are used to attempt to model the complex correspondences between visual and auditory features. The result of this project is a software tool allowing users to simply define pairs of sounds and images, which can then be extrapolated to generate images for new sounds.