Proposing the Application of LSTM Models to Create Music

• Published in: 2024 4th International Conference on Electrical Engineering and Informatics (ICon EEI) pp. 102 - 107
• Main Authors: Tan, Nghia Duong, Anh, Bang Nguyen, Thanh, Xuyen Do Bui, Phuc, Lam Nguyen Tan
• Format: Conference Proceeding
• Language: English
• Published: IEEE 16-10-2024
• Subjects: Artificial Intelligence; Data models; Deep learning; Entertainment industry; Feature extraction; Long short term memory; Music; Music generation; Neural networks; Optimization; Quality of experience; Recurrent neural network; Training; Training data
• DOI: 10.1109/IConEEI64414.2024.10747903

In recent years, artificial intelligence (AI) and more specifically machine learning (ML) have emerged as a demonstration of the 4th industrial revolution. AI is present in all areas of human life from economics, education, and healthcare to housework, entertainment, and even the military. Outstanding applications in AI development come from many fields to solve different problems of life. The paper aims to learn and apply machine learning in sound processing for entertainment. In the paper, we use the long short-term memory (LSTM) model to create music. To perform the model, we have to collect data and preprocess to standardize the model training data. We then build a deep learning (DL) model using the LSTM network and train it with optimization parameters to create new, unique, and creative music. As a result, an automatic music generation model does not require human intervention and can generate highly rated music tracks based on the QoE factor as 5.15; 3.65; and 3.7 points for tracks 1, 2, and 3, respectively. The results show that the AI model has the potential to be applied in many areas of life.