Combining pitch-based inference and non-negative spectrogram factorization in separating vocals from polyphonic music.

Tuomas Virtanen, Annamaria Mesaros, Matti Ryynnen

Browse the full Interspeech paper archive.