Music Tempo (Speed) Classification
Yu-Yao Chang, Yao-Chung Lin · 2005
Music tempo (speed) is one of the most important features of a song. With successful classification of the tempo of song, content-based music browsing may utilize this feature and search/recommend songs in the same category of tempo. With some paper survey we noticed that prior works were focused on the numerical tempo value instead of the human perceptual speed of song. We proposed Inter-Onset Interval (IOI) reliability feature to present a soft tempo information for SVM. The onset detection algorithm in [4] was implemented to extract the beat onset timing, and the detection result is applied to SVM [6] as training features. The model trained via SVM can classify slow and medium songs with high accuracy; however, the fast songs are not highly correctly classified. We consider improving the feature for fast songs as future work.