A Perceptual Measure for Evaluating the Resynthesis of Automatic Music Transcriptions
This work addresses the gap in assessing artistic intention in music performance for researchers and developers in music technology, though it is incremental as it builds on existing transcription methods.
The study tackled the problem of evaluating how well automatic music transcription systems capture artistic intention by testing 91 subjects' perceptions of resynthesized music under different contextual changes, finding that MIDI data alone cannot fully capture artistic intention and that existing objective measures poorly correlate with subjective evaluations.
This study focuses on the perception of music performances when contextual factors, such as room acoustics and instrument, change. We propose to distinguish the concept of "performance" from the one of "interpretation", which expresses the "artistic intention". Towards assessing this distinction, we carried out an experimental evaluation where 91 subjects were invited to listen to various audio recordings created by resynthesizing MIDI data obtained through Automatic Music Transcription (AMT) systems and a sensorized acoustic piano. During the resynthesis, we simulated different contexts and asked listeners to evaluate how much the interpretation changes when the context changes. Results show that: (1) MIDI format alone is not able to completely grasp the artistic intention of a music performance; (2) usual objective evaluation measures based on MIDI data present low correlations with the average subjective evaluation. To bridge this gap, we propose a novel measure which is meaningfully correlated with the outcome of the tests. In addition, we investigate multimodal machine learning by providing a new score-informed AMT method and propose an approximation algorithm for the $p$-dispersion problem.