Source separation is the process of separating (or un-mixing) the source components in a polyphonic and possibly multitimbral audio signal.
Source separation is a discipline within Digital Signal Processing concerned with the isolation (and possible removal) of source components of a complex signals such as polyphonic and multi-timbral audio signals.
The source separation process can be considered to be the inverse of mixing. However, the process of mixing signals loses information, and a source separation algorithm attempts to reconstruct the missing parts. For this reason, the process is never perfect, and artefacts result.
A common commercial application of audio source separaton is the removal of vocals from pop-songs in order to create karaoke backings. It is also an important tool used in Music Information Retrieval (MIR). Audio source separation is a subject of considerable current academic interest.