Related Experiment Videos
Data processing in music performance research: using structural information to improve score-performance matching
H Heijink1, L Windsor, P Desain
1NICI, University of Nijmegen, P. O. Box 9104, 6500 HE Nijmegen, The Netherlands. heijink@nici.kun.nl
Behavior Research Methods, Instruments, & Computers : a Journal of the Psychonomic Society, Inc
|February 24, 2001
Summary
Automated music performance matching algorithms are improved by incorporating musical structure. However, this alone cannot fully identify performance errors or define the best match, necessitating further algorithmic development.
Area of Science:
- Music Performance Analysis
- Computational Musicology
- Algorithmic Musicology
Background:
- Automated matching of music performance data to a score is crucial for analysis.
- Current algorithms face challenges due to performance errors, expressive timing, and score underspecification.
Purpose of the Study:
- To evaluate the impact of incorporating musical structure into automated music performance matching algorithms.
- To identify limitations of current matching approaches and suggest areas for improvement.
Main Methods:
- Developed and tested a music performance matching algorithm incorporating score musical structure.
- Compared performance against traditional matchers relying on pitch, temporal order, and note count.
Main Results:
- Integrating musical structure information significantly improved matching accuracy compared to traditional methods.
- Even with structural information, identifying certain performance errors and defining the 'best match' remained problematic.
Conclusions:
- Musical structure is a valuable feature for enhancing automated music performance matching.
- Further research is needed to address limitations in error detection and best match definition for more robust algorithms.