期刊名称：International Journal of Advanced Research in Computer Engineering & Technology (IJARCET)
印刷版ISSN：2278-1323
出版年度：2015
卷号：4
期号：2
页码：358-361
出版社：Shri Pannalal Research Institute of Technolgy
摘要：Extraction of singing voice from music signal is required for many applications such as MIR (Music Information Retrieval), automatic karaoke generator, interactive music player etc.. In many MIR studies like automatic lyrics recognition, identification of language of a song, automatic singer identification etc., has used the information on singing voice. But singing voice separation is a challenging problem. Since both singing voice and musical sounds are harmonic, simple harmonic extraction technique should not be used. And music signals do not satisfy the properties of noise. So classical noise suppression techniques will not work effectively. This paper is about various methods to separate singing voice from monaural music signals. Some technique uses models for singing voice separation; some others use properties of singing voice like high rankness, fluctuation etc.
关键词：singing voice separation; Music Information ; Retrieval; harmonic; noise; low rankness; Fluctuation
