Evaluating automatic melody segmentation aimed at music information retrieval

Massimo Melucci, Nicola Orio
2002 Proceedings of the second ACM/IEEE-CS joint conference on Digital libraries - JCDL '02  
In this paper we investigate the effectiveness of a melody segmentation algorithm based on melodic feature. The segmentation produced by experienced music scholars have been compared with the algorithm, a random segmenter and a ngram-based segmenter. Results showed that the algorithm is closer to manual segmentation than the other segmenters.
doi:10.1145/544220.544294 dblp:conf/jcdl/MelucciO02 fatcat:npqzwijpq5fnxc3vrcg3zk4vd4