Open AccessA MUSIC GENERATION BY A COMBINING MODELOF RESNET AND LSTM NETWORKS
Author(s) -
Kazuya Ozawa,
Hiroyuki Okazaki
Publication year - 2022
Publication title -
international journal of advanced research
Language(s) - English
Resource type - Journals
ISSN - 2320-5407
DOI - 10.21474/ijar01/14474
Subject(s) - computer science , piano , long short term memory , residual neural network , speech recognition , residual , artificial intelligence , deep learning , artificial neural network , recurrent neural network , machine learning , algorithm , art , art history
In this paper, to automatically generate a music for the melody part by deep learning with training data collected from Chopins piano piecies, a combining model of Residual Neural Networks(ResNet) and Long-Short Term Memory Networks (LSTM) are proposed. First, to generate a music for the melody part of a piano music, a training dataset used for deep learning is provided. Secondly, by using each of a LSTM Model and a combining model of LSTM and ResNet,experiments on music generationare presented. Thirdly, the results of music generation by each model are compared and discussed. In conclusion, the principal results are summarized.