The Application of Long Short-Term Memory and Transformers for Music Generation

Publication Date

1-1-2023

Document Type

Conference Proceeding

Publication Title

Proceedings - 2023 IEEE International Conference on Big Data, BigData 2023

DOI

10.1109/BigData59044.2023.10386572

First Page

4475

Last Page

4478

Abstract

Music is embedded deeply in our well-being, and it has a great impact on improving mood, decreasing pain and anxiety, and providing opportunities to express oneself emotionally. Upon hearing a familiar tune, the human body releases a neurotransmitter called Dopamine which is responsible for the feeling of pleasure and comfort. Research suggests that music helps the brain cells to process information more effectively. Studies prove that people who suffered from stroke gained back their cognition within two months by listening to music when compared to the ones who did not listen to music. Music comprises numerous elements like pitch, rhythm, tone, etc. Music is subjective and broadly classified by genre. Music is also shown to have health benefits, both physically and emotionally. With advancements in generative AI, there is a unique opportunity to supply new and fresh tunes using knowledge of the type of music a patient listens to. In this paper, we present a deep-learning technique to create music samples. We experiment with Long Short-Term Memory (LSTM) and transformers in generating music. We envision that our platform will generate novel music samples of a particular genre and in the long-term help patients in need of physical and emotional strength.

Keywords

BERT, Deep Learning, Generative Adversarial Networks (GANs), Genre, Long Short-Term Memory (LSTM), Recurrent Neural Networks (RNNs), Transformers, Variational Auto Encoders (VAEs)

Department

Computer Engineering

Link to Full Text

Share

COinS