A Novel Extraction Method for Melodic Features from MIDI Files Based on Probabilistic Graphical Models

This paper, using MIDI file as the research object, presents a naïve bayes classifier and probabilistic graphical model (PGM), which is designed by characterizing the extraction of the melody vectors of the music features from each track of the MIDI file, and the MIDI melody tracks and accompaniment melody tracks are automatically classified. Finally, through the candidate audio tracks extracted from the main melody track. This method does not require a priori knowledge of music. Evaluation shows that when compared with other methods, the proposed approach is outperformed in recognition precision and is effective for solving the melody recognition problem.