Skip That Beat: Augmenting Meter Tracking Models for Underrepresented Time Signatures

TL;DR

This paper introduces a data augmentation method that enhances beat and downbeat tracking models' ability to recognize less common time signatures like 2/4 and 3/4, improving performance on diverse musical styles.

Contribution

The authors propose a simple augmentation technique that increases the representation of underrepresented time signatures, improving model generalization without sacrificing overall accuracy.

Findings

Improved downbeat tracking for 2/4 and 3/4 signatures.

Enhanced performance on unseen samba dataset.

Preserved beat tracking accuracy in models.

Abstract

Beat and downbeat tracking models are predominantly developed using datasets with music in 4/4 meter, which decreases their generalization to repertories in other time signatures, such as Brazilian samba which is in 2/4. In this work, we propose a simple augmentation technique to increase the representation of time signatures beyond 4/4, namely 2/4 and 3/4. Our augmentation procedure works by removing beat intervals from 4/4 annotated tracks. We show that the augmented data helps to improve downbeat tracking for underrepresented meters while preserving the overall performance of beat tracking in two different models. We also show that this technique helps improve downbeat tracking in an unseen samba dataset.