Proceeding of the past Conferences /

Below you can read the papers of the past Conferences

New Music Concepts 2020

Conditional Modelling of Musical Bars with Convolutional Variational Autoencoder

Part of the New Music Concepts book series (NMC volume 7)
ISBN: 978-88-944350-3-0

Author(s): A. Oudad, H. Saito

Abstract: Generational machine learning has been applied to various fields and is currently under exploration for music. Automatic music composition is receiving attention recently thanks to advance in distributed modelling using neural networks. Yet, musical models lack the ability to effectively capture harmonic and rhythmic features of music. We propose a model using variational autoencoder for encoding musical bars of ragtime piano music. Ragtime piano music is composed of left hand and right hand with different roles, thus we use righthand musical information to condition our variational autoencoder when reconstructing left-hand musical bars. We show our model is able to reconstruct and interpolate musical bars, thus providing useful musical bar embeddings for music generation.

Keywords: Deep Learning, Algorithmic Music Composition, Variational Inference, Music Information Processing