Audio segmentation based on multi-scale audio classification
Yibin Zhang, Jie Zhou · 2004
Content-based audio segmentation plays an important role in multimedia applications. In order to segment accurately and on-line, most conventional algorithms are based on small-scale feature classification and always result in a high false alarm rate. Our experimental results show that large-scale audio can be more easily classified than small ones. According to this fact, we present a novel multi-scale framework for audio segmentation. First, a rough segmentation step based on large-scale classification is taken to ensure the integrality of the content of segments, which can avoid the consecutive audio belonging to the same kind being segmented into different pieces. Then a subtle segmentation step is taken to further locate the segmentation points for the boundary areas computed by the rough segmentation step. Experimental results show that a low false alarm rate can be achieved while preserving a low missing rate.