Automatic summarization for popular songs

In this paper, we present an automatic summarization approach for popular songs. The approach includes two stages. First, a rough summary of the song is extracted efficiently by the energy information and the songpsilas structure, and then the segment boundaries of the summary are detected. Different from most previous works, the summary has a more complete meaning and a varied length instead of a fixed length, therefore it is much helpful for popular songpsilas audition and management. The algorithm is tested and evaluated by objective means. Experimental results show that the proposed approach achieves satisfactory results.