Separating Advertisements and DJ Chatter from Music

Music on the radio is often interrupted by advertisements or DJ chatter that the listener would prefer not to hear. In this paper, we discuss a straightforward application of perceptually-based audio feature extraction and classification using a support vector machine to automatically differentiate between music and “non-music” audio, so that when non-music is detected the radio might automatically seek out more music on other stations. Time F re q u e n c y 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 0 0.5 1 1.5 2 x 10 4 Samples F e a tu re s 20 40 60 80 100 120 140 160 180 5