Abstract

This extended abstract details a submission to the Music Information Retrieval Evaluation eXchange in the Audio Genre classification task. This submission is very similar to the system that placed second in the 2004 ISMIR Audio description contest. A novel feature set and segmentation of features is introduced and modifications to the Decision Tree based model used in the 2004 submission are detailed. Finally, the results achieved in the evaluation are analysed. Keywords: MIREX, Audio, Genre. 1 FEATURE SET Two feature sets are calculated in this submission, one describing the timbre of the audio and another describing the rhythmic content. 22 kHz audio is taken as input,

Citations

27	Music type classification by spectral contrast feature - Jiang, Lu, et al. - 2002
15	Finding an optimal segmentation for audio genre classification - West, Cox
4	Brooks/Cole Advanced Books & Software - Wadsworth - 1986