The beat spectrum: a new approach to rhythm analysis

Abstract

We introduce the beat spectrum, a new method of automatically characterizing the rhythm and tempo of music and audio. The beat spectrum is a measure of acoustic self-similarity as a function of time lag. Highly structured or repetitive music will have strong beat spectrum peaks at the repetition times. This reveals both tempo and the relative strength of particular beats, and therefore can distinguish between different kinds of rhythms at the same tempo. We also introduce the beat spectrogram which graphically illustrates rhythm variation over time. Unlike previous approaches to tempo analysis, the beat spectrum does not depend on particular attributes such as energy or frequency, and thus will work for any music or audio in any genre. We present tempo estimation results which are accurate to within 1% for a variety of musical genres. This approach has a variety of applications, including music retrieval by similarity and automatically generating music videos. Anyone who has ever tapped a foot in time to music has performed rhythm analysis. Though simple for humans, this task is considerably more difficult to automate. We introduce a new measure of tempo analysis called the beat spectrum. This is a measure of acoustic self-similarity versus lag time, computed from a representation of spectrally similarity. Peaks in the beat spectrum correspond to major rhythmic components of the source audio. The repetition time of each component can be determined by the lag time of the corresponding peak, while the relative amplitudes of different peaks reflects the strengths of their corresponding rhythmic components. We also present the beat spectrogram which graphically illustrates rhythmic variation over time. The beat spectrogram is an image formed from the beat spectrum over successive windows. Strong rhythmic components are visible as bright bars in the beat spectrogram, making changes in tempo or time signature visible. In addition, a measure of audio novelty can be computed that measures how novel the source audio is at any time [2]. Instances when this measure is large correspond to significant audio changes. Periodic peaks correspond to rhythmic periodicity in the music. In the final section, we present various applications of the beat spectrum, including music retrieval by rhythmic similarity, an “automatic DJ” that can smoothly sequence music with similar tempos and automatic music video generation.