Rhythm Classification Using Spectral Rhythm Patterns

In this paper, we study the use of spectral patterns to represent the characteristics of the rhythm of an audio signal. A function representing the position of onsets over time is first extracted from the audio signal. From this function we compute at each time a vector which represents the characteristics of the local rhythm. Three feature sets are studied for this vector. They are derived from the amplitude of the Discrete Fourier Transform, the AutoCorrelation Function and the product of the DFT and of a Frequency-Mapped ACF. The vectors are then sampled at some specific frequencies, which represents various ratios of the local tempo. The ability of the three feature sets to represent the rhythm characteristics of an audio item is evaluated through a classification task. We show that using such simple spectral representations allows obtaining results comparable to the state of the art.