Open AccessArtificial Music Generation using LSTM Networks
Author(s) -
Hemalatha Eedi
Publication year - 2019
Publication title -
international journal of engineering and advanced technology
Language(s) - English
Resource type - Journals
ISSN - 2249-8958
DOI - 10.35940/ijeat.b4522.129219
Subject(s) - melody , computer science , artificial neural network , measure (data warehouse) , layer (electronics) , artificial intelligence , sample (material) , variation (astronomy) , musical composition , software , machine learning , data mining , music education , art , musical , chemistry , physics , organic chemistry , chromatography , astrophysics , visual arts , programming language
Advancements in machine learning have minimized the gap of variation between human and algorithm composed music. This paper realizes a music generation system using evolutionary algorithms. The music generation is fully automated with no requirement of human intervention. Multiple music sample from a single dataset were used to the neural network. Software has been constructed to exhibit the results over various datasets. The proposed model is based on recurrent neural network with the input layer represents a measure at time T, and the output layer represents the measure at time T+1. The approach results in generation of new music composition by the system. Composition rules are used as constraints to evaluate the melodies generated by the novel neural network. Thus, the results are expected to evolve to satisfy the defined constraints. The proposed system of work would be capable of music generation without human intervention.