A Lightweight Deep Learning-Based Approach for Jazz Music Generation in MIDI Format.

Area of Science:

• Artificial Intelligence
• Music Information Retrieval
• Computational Musicology

Background:

• Accurate musical difficulty estimation is crucial for effective music learning but is complicated by subjective content and data scarcity.
• Existing music generation models often lack genre specificity or require extensive computational resources.

Purpose of the Study:

• To propose a lightweight deep learning model for generating original jazz music in MIDI format.
• To enable music generation based on specific genres and input musical segments.
• To address the challenges of subjectivity and data scarcity in computational music generation.

Main Methods:

• Developed a lightweight deep learning model trained on classical jazz music in MIDI format.
• The model takes a segment of music as input and generates its continuation.
• Ensured dataset homogeneity to maintain desired output characteristics, focusing on classical jazz pieces.

Main Results:

• The model successfully generates novel jazz music continuations in MIDI format.
• Generated music exhibits similarity to the input data and adheres to the jazz genre.
• The model demonstrates the capability to generate music with specific instrumental characteristics.

Conclusions:

• The proposed lightweight deep learning approach offers an effective method for generating genre-specific music, particularly jazz.
• This model can aid in music learning by providing tools for content generation and analysis.
• Future work could explore broader genre applications and more sophisticated control over generated musical elements.