Procedural Composition of Traditional Persian Music Using Deep Neural Networks

Abstract

In recent years the explosion in the amount of digital content used in various forms in virtual reality environments such as videogames requires generating a massive amount of new artistic materials. Procedural content generation using artificial intelligence algorithms helps producing this artistic materials with significant variety and based on various traditional source materials and avoiding virtual environments to be derivative. These systems can also lead to multi-cultural variety and use of traditional art in digital environments. In this paper, a recurrent deep neural network based framework for procedural composition of the traditional Persian music is proposed. The proposed system learns from the classical Persian musical modal systems and then produces new music. The proposed system is evaluated and the results show that the deep neural networks are capable of producing traditional Persian music with acceptable quality.