Manni CHEN, PerMagnus LINDBORG
Abstract:
Generating music with pitch and rhythm has received ample attention from researchers. By contrast, noise music as a special genre has been largely ignored so far. This paper aims to generate noise music in ‘Merzbow style’ using a neural network called variational autoencoder (VAE). The result shows that our proof-of-concept system is capable of generating noise music. Examples of output are audio samples that can be listened on our website.
The neural network structure:
Audio samples can be listened here: