A novel approach to automatic singer identification in duet recordings with background accompaniments

The musical attribute that mostly attracts the listener's attention is the singing voice. The singer's identity serves as a primary aid for people to organize, browse, and retrieve music recordings. Major problems faced by (Singer Identification) SID techniques are that, they will have to deal with music recordings having background accompaniments, requirement of solo or cappella data of all singers which is practically not viable, efficient segmentation of vocal and non-vocal parts in a recording and it will also have to deal with identifying the singer in solo music recordings with background accompaniment. Considering all these challenges a novel method is proposed here, which deals with identifying the vocal parts in the audio signal and removal of background accompaniments. This work also deals with identification of singers in duet recordings using singer specific models.