Sound and Music Computing Seminar / Master of Science in Music and Technology Thesis Defense
9115 - Gates Hillman Centers
ZHENG JIANG , Masters Student, Music and Technology
Computer Based Music Analysis
The identification of melody, bass, chord and structure are important steps in music analysis. This thesis presents a comprehensive tool to identify the melody and bass in each measure of a Standard MIDI File, and provides chord labels and general structure information. We also share an open dataset of manually labeled music for researchers. We use a Bayesian maximum-likelihood approach and dynamic programming as the basis of our work in the melody identification. We have trained parameters on data sampled from the million song dataset and tested on a dataset including 1703 measures of music from different genres. Our algorithm achieves an overall accuracy of 89% in the test dataset. We compare our results to previous work. For bass identification, since our algorithm is rule-based, we only label the test set. And the bass identification achieves over 95% accuracy. Our chord-labeling algorithm is adopted from Temperley and tested on a manually labeled test set containing 1890 labels. Currently, this algorithm achieves around 78% accuracy for chord root and chord type matching and around 82% accuracy for chord root matching only. We also discovered an optimization: by removing notes in melody channel, we can improve the accuracy for both evaluations by 1.3%. For the structure analysis, we also provide a simple algorithm based on a similarity matrix and give some analysis of the result. By automatically labeling and analyzing MIDI files, a rich source of symbolic music information, we hope to enable new studies of music style, music composition, music structure, and computational music theory.
Master of Science in Music and Technology Thesis Defense
For More Information: