A tutorial on onset detection in music signals

TLDR

Onset detection and localization are essential for musical signal analysis and indexing, typically achieved by identifying transient regions defined by sudden energy bursts, spectral changes, or statistical property shifts. The paper reviews, categorizes, and compares common onset detection techniques and proposes possible enhancements. The authors examine feature‑based methods using amplitude envelopes, spectral magnitudes, phases, and time‑frequency representations, as well as probabilistic models such as change‑point detection and surprise signals. The study offers guidelines, derived from test cases, for selecting the most suitable onset detection method for a specific application.

Abstract

Note onset detection and localization is useful in a number of analysis and indexing techniques for musical signals. The usual way to detect onsets is to look for "transient" regions in the signal, a notion that leads to many definitions: a sudden burst of energy, a change in the short-time spectrum of the signal or in the statistical properties, etc. The goal of this paper is to review, categorize, and compare some of the most commonly used techniques for onset detection, and to present possible enhancements. We discuss methods based on the use of explicitly predefined signal features: the signal's amplitude envelope, spectral magnitudes and phases, time-frequency representations; and methods based on probabilistic signal models: model-based change point detection, surprise signals, etc. Using a choice of test cases, we provide some guidelines for choosing the appropriate method for a given application.