Singer's voice elimination from stereophonic pop music using ICA

Separation of vocal elements from background music is a popular application in music information retrieval. Additionally, it is known that if two independent and non-Gaussian signals are mixed together, independent component analysis can be employed to separate the two sources from their mixtures. For this reason, in this paper, a fast and simple approach based on independent component analysis has been proposed to separate the singer's voice from the stereophonic recorded music. Finally, the proposed algorithm has been evaluated using different pieces of pop music.