Rethinking Evaluation Methodology for Audio-to-Score Alignment

TL;DR

This paper provides a formal definition of audio-to-score alignment, introduces new evaluation metrics, and analyzes their effectiveness on classical algorithms using a specialized dataset.

Contribution

It offers a precise formalization of audio-to-score alignment and proposes novel evaluation metrics to improve assessment accuracy.

Findings

New metrics reveal different algorithm performances

Classical algorithms show varied results with new metrics

Formal definition enhances alignment evaluation clarity

Abstract

This paper offers a precise, formal definition of an audio-to-score alignment. While the concept of an alignment is intuitively grasped, this precision affords us new insight into the evaluation of audio-to-score alignment algorithms. Motivated by these insights, we introduce new evaluation metrics for audio-to-score alignment. Using an alignment evaluation dataset derived from pairs of KernScores and MAESTRO performances, we study the behavior of our new metrics and the standard metrics on several classical alignment algorithms.