Singing information processing: Concept and applications.
Masataka Goto, Takeshi Saitou, Tomoyasu Nakano, Hiromasa Fujihara · The Journal of the Acoustical Society of America · 2010
This paper describes recent research on singing information processing [Goto et al., Proc. of IEEE ICASSP (2010)], which is music information research for singing voices. Singing is one of the most important elements of music since a great number of people listen to music with a focus on singing, especially in the case of popular music. The concept of singing information processing systems is broad and still emerging, but three important categories are singing understanding systems, music information retrieval systems based on singing voices, and singing synthesis systems. Singing understanding systems have been developed for various tasks such as synchronizing lyrics to vocal music, singer name identification, singing skill evaluation, creating hyperlinks between phrases in the lyrics of songs, and detecting breath sounds. Music information retrieval systems based on similarity of vocal melody timbre and vocal percussion, as well as singing synthesis systems for speech-to-singing synthesis and singing-to-singing synthesis, have also been developed. Common signal processing techniques, such as techniques for extracting vocal melody from polyphonic music recordings and modeling the lyrics by using phoneme hidden Markov models for singing voices, are used in these systems. [This research was supported in part by CrestMuse, CREST, JST.]