Automatic audio segmentation using a measure of audio novelty

TLDR

The paper proposes a method to automatically locate significant change points in music or audio by analyzing local self‑similarity. The method models the signal itself using local self‑similarity, avoiding reliance on specific acoustic cues or training data. The approach accurately detects note, verse/chorus, and speech/music boundaries, supports indexing, segmentation, and beat tracking, and performs well across diverse audio sources.

Abstract

The paper describes methods for automatically locating points of significant change in music or audio, by analyzing local self-similarity. This method can find individual note boundaries or even natural segment boundaries such as verse/chorus or speech/music transitions, even in the absence of cues such as silence. This approach uses the signal to model itself, and thus does not rely on particular acoustic cues nor requires training. We present a wide variety of applications, including indexing, segmenting, and beat tracking of music and audio. The method works well on a wide variety of audio sources.