Automatic Identification of Simultaneous Singers in Duet Recordings

The problem of identifying singers in music recordings has received considerable attention with the explosive growth of the Internet and digital media. Although a number of studies on automatic singer identification from acoustic features have been reported, most systems to date, however, reliably establish the identity of singers in solo recordings only. The research presented in this paper attempts to automatically identify singers in music recordings that contain overlapping singing voices. Two approaches to overlapping singer identification are proposed and evaluated. Results obtained demonstrate the feasibility of the systems.