MIR on Representative Mexican Folk Vocal Melodies Through MIDI Feature Extraction

I'm pleased to share my latest research paper, "Music Information Retrieval on Representative Mexican Folk Vocal Melodies Through MIDI Feature Extraction," now available on arXiv. This study analyzes Mexican folk vocal melodies, focusing on MIDI-derived features such as ambitus, pitch-class entropy, interval distribution, and their potential relationship with song popularity measured by Spotify play counts.

Key Findings:

• Ambitus varied from 8 to 27 semitones, highlighting diverse vocal demands and compositional styles within Mexican folk music.

• Pitch-class entropy revealed varied melodic complexity, with Armando Manzanero's "Somos Novios" exhibiting the highest entropy (0.85), suggesting an intricate melodic structure. Traditional pieces like "La Bamba" displayed lower entropy (0.68), indicating simpler melodic patterns.

• Interval distribution predominantly featured prime intervals (P1), major seconds (M2), and minor seconds (m2), emphasizing a preference for close intervals contributing to melodic accessibility.

• Statistical analyses showed no significant correlation between ambitus or pitch-class entropy and Spotify play counts, suggesting melodic complexity alone does not predict popularity.
  

Methodology: The research used a Music Information Retrieval (MIR) pipeline developed with MATLAB and the MIDI Toolbox. Vocal melodies from MIDI files of culturally significant Mexican songs were isolated and normalized using Pro Tools. MATLAB was used for feature extraction and statistical evaluation using the Pearson correlation coefficient.

Implications: This study provides insights into Mexican musicology and computational music analysis, enhancing understanding of melodic structures in culturally influential songs. The lack of significant correlation with Spotify plays underscores the multifaceted nature of music popularity, involving factors beyond musical attributes.

Full Paper and Resources: The complete research paper is available on arXiv.

For further discussion or inquiries, please feel free to reach out directly.