Metrical-accent Aware Vocal Onset Detection in Polyphonic Audio

TL;DR

This paper introduces a probabilistic model that leverages metrical accents to improve vocal onset detection accuracy in polyphonic music, validated across diverse musical styles and traditions.

Contribution

A novel probabilistic approach that jointly tracks beats and vocal onsets using metrical information, enhancing detection accuracy over existing models.

Findings

Improved onset detection accuracy with metrical awareness

Effective across multiple musical genres and traditions

Validates the benefit of incorporating metrical context

Abstract

The goal of this study is the automatic detection of onsets of the singing voice in polyphonic audio recordings. Starting with a hypothesis that the knowledge of the current position in a metrical cycle (i.e. metrical accent) can improve the accuracy of vocal note onset detection, we propose a novel probabilistic model to jointly track beats and vocal note onsets. The proposed model extends a state of the art model for beat and meter tracking, in which a-priori probability of a note at a specific metrical accent interacts with the probability of observing a vocal note onset. We carry out an evaluation on a varied collection of multi-instrument datasets from two music traditions (English popular music and Turkish makam) with different types of metrical cycles and singing styles. Results confirm that the proposed model reasonably improves vocal note onset detection accuracy compared to a baseline model that does not take metrical position into account.