Abstract
Tempogram is one of the most useful representations for tempo, which has many applications, such as music tempo estimation, music structure analysis, music classification, and beat tracking. This paper presents a novel tempogram generating algorithm, which is based on matching pursuit. First, a tempo dictionary is designed in the light of the characteristics of tempo and note onset, then matching pursuit based on the tempo dictionary is executed on the resampled novelty curve, and finally the tempogram is created by assembling the coefficients of matching pursuit. The tempogram created by this algorithm has better resolution, stronger sparsity, and flexibility than those of the traditional algorithms. We demonstrate the properties of the algorithm through experiments and provide an application example for tempo estimation.
Highlights
IntroductionTempo is the speed or pace of a given piece [1]
In musical terminology, tempo is the speed or pace of a given piece [1]
To compare the tempo resolution, we show three tempograms of the clip in the Figure 4, in which the tempogram of Figure 4a–c are created by the methods of autocorrelation function (ACF), Fourier transform (FT), and our MP, respectively
Summary
Tempo is the speed or pace of a given piece [1]. It is usually measured by beats per minute (bpm). In 4/4 the beat will be a crotchet or quarter note. Tempo is one of the most important features of music, which is closely relevant to beat and rhythm. Tempo estimation is one of the most important research fields in music information retrieval (MIR), and it is the fundamental work for many other applications, such as beat tracking [2,3,4], music structure analysis [5], rhythm identification [6], and music classification [7]
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
