tonset

Tonset is a term used in acoustics and music information retrieval to denote the moment at which a tonal sound begins within an audio signal. It marks the onset of a tone, often corresponding to the attack of a note, and is used to distinguish the initial moment from the subsequent sustain or decay. In practice, tonset detection algorithms aim to identify these moments across a recording to support higher-level analyses like transcription, tempo estimation, and instrument recognition.

Etymology and usage: The word tonset is a blend of tone and onset. In some contexts it is used interchangeably with note onset or onset when the focus is on tonal content rather than percussive events, though distinctions may be made in specialized literature.

Methods: Onset detection relies on features that reveal sudden changes in the signal. Common approaches include energy-based methods that detect rapid increases in short-time energy, spectral flux measures that track frame-to-frame spectral changes, and phase-based techniques that exploit phase coherence. More robust systems often combine multiple features or use supervised learning to classify tonsets in polyphonic material.

Applications: Tonset information feeds automatic music transcription, rhythm and beat tracking, segmentation of audio streams, and real-time music analysis in interactive systems.

History and reception: Onset concepts have been studied since the early days of digital signal processing, and tonset detection remains a fundamental preprocessing step in many music information retrieval pipelines.

See also: onset, attack (music), note onset time, beat tracking.