Representative Music Fragments Extraction by Using Segmentation Techniques

Abstract

Melody extraction is an important issue of research for successful development of music information retrieval system, particularly for polyphonic music. The extracted melody can further processed as music index that can speed up the retrieval of large music collections. In the past, most researchers assume music is monophonic. If the music is polyphonic, there is not satisfyingly solved by existing known algorithms. In this paper, we propose two methods: “Four-Measure Based segmentation” and “LBDM Concatenation” to extract the representative fragments of melody from di®erent tracks. We also analyze the e®ects caused by the varied combinations of music features to help us extract the melody.