Abstract

In this paper we present beat histogram features for multiple level rhythmdescriptionandevaluatetheminamusicalgenreclassifica- tion task. Audio features pertaining to various musical content cat- egories and their related novelty functions are extracted as a basis for the creation of beat histograms. The proposed features capture not only amplitude, but also tonal and general spectral changes in the signal, aiming to represent as much rhythmic information as possible. The most and least informative features are identi- fied through feature selection methods and are then tested using Support Vector Machines on five genre datasets concerning classi- fication accuracy against a baseline feature set. Results show that the presented features provide comparable classification accuracy with respect to other genre classification approaches using period- icity histograms and display a performance close to that of much more elaborate up-to-date approaches for rhythm description. The use of bar boundary annotations for the texture frames has pro- vided an improvement for the dance-oriented Ballroom dataset. The comparably small number of descriptors and the possibility of evaluating the influence of specific signal components to the gen- eral rhythmic content encourage the further use of the method in rhythm description tasks.