Estimating tempo, swing and beat locations in audio recordings

The problem of estimating the tempo of audio recordings (the number of beats per minute, or BPM) has received an increasing amount of attention in the past few years. Applications include the synchronization of multiple audio tracks for simultaneous playback, "tempo-synchronous" audio effects, automatic looping of audio tracks etc. This article presents techniques for estimating the tempo and the swing, and locating the beats in audio recordings, under the assumption that the tempo is constant. The techniques rely on a preliminary transient detection stage where note onsets/offsets, percussion hits and other time-localized events are detected. This first step is followed by a maximum likelihood estimation of the tempo, swing and downbeat. Suggestions are given to minimize the computation load of the methods