Publicação
Deep Learning for Dynamic Music Generation
| Resumo: | In the last decade, Deep Learning (DL) algorithms have been increasing their popularity in several fields such as computer vision, speech recognition, natural language processing and many others. DL models, however, are not limited to scientific domains as they have recently been applied to content generation in diverse art forms - both in the generation of novel content and as co-creative tools. Artificial music generation is one of the fields where DL architectures have been applied. They have been mostly used to create new compositions exhibiting promising results when compared to human compositions. Despite this, the majority of these artificial pieces lack some expression when compared to music compositions performed by humans. In this document, we propose a system capable of artificially generating expressive music compositions. Our main goal is to improve the quality of the musical compositions generated by the artificial system by exploring perceptually relevant musical elements such as note velocity and duration. In a primary analysis, the expressive compositions generated by our model present expressive variations that improve the dynamics of the piece, which can be verified by comparing non-expressive pieces with expressive ones (either in humans or in artificial compositions generated by our model).To assess this hypothesis we perform user tests. Our results suggest that expressive elements such as duration and velocity are key aspects in a music composition, making the ones that include these preferable to non-expressive ones. |
|---|---|
| Autores principais: | Simões, José Maria da Costa |
| Assunto: | Deep Learning Recurrent Neural Networks Composição Musical Duração Velocidade Deep Learning Recurrent Neural Networks Music Composition Duration Velocity |
| Ano: | 2019 |
| País: | Portugal |
| Tipo de documento: | dissertação de mestrado |
| Tipo de acesso: | acesso aberto |
| Instituição associada: | Universidade de Coimbra |
| Idioma: | inglês |
| Origem: | Estudo Geral - Universidade de Coimbra |
| Resumo: | In the last decade, Deep Learning (DL) algorithms have been increasing their popularity in several fields such as computer vision, speech recognition, natural language processing and many others. DL models, however, are not limited to scientific domains as they have recently been applied to content generation in diverse art forms - both in the generation of novel content and as co-creative tools. Artificial music generation is one of the fields where DL architectures have been applied. They have been mostly used to create new compositions exhibiting promising results when compared to human compositions. Despite this, the majority of these artificial pieces lack some expression when compared to music compositions performed by humans. In this document, we propose a system capable of artificially generating expressive music compositions. Our main goal is to improve the quality of the musical compositions generated by the artificial system by exploring perceptually relevant musical elements such as note velocity and duration. In a primary analysis, the expressive compositions generated by our model present expressive variations that improve the dynamics of the piece, which can be verified by comparing non-expressive pieces with expressive ones (either in humans or in artificial compositions generated by our model).To assess this hypothesis we perform user tests. Our results suggest that expressive elements such as duration and velocity are key aspects in a music composition, making the ones that include these preferable to non-expressive ones. |
|---|