Oracle estimators for the benchmarking of source separation algorithms
Signal Processing
Blind separation of speech mixtures via time-frequency masking
IEEE Transactions on Signal Processing
IEEE Transactions on Information Theory
IEEE Transactions on Audio, Speech, and Language Processing
Hi-index | 0.00 |
With a standard compact disc (CD) audio player, the only possibility for the user is to listen to the recorded track, passively: the interaction is limited to changing the global volume or the track. Imagine now that the listener can turn into a musician, playing with the sound sources present in the stereo mix, changing their respective volumes and locations in space. For example, a given instrument or voice can be either muted, amplified, or more generally moved in the acoustic space. This will be a kind of generalized karaoke, useful for disc jockeys and also for music pedagogy (when practicing an instrument). Our system shows that this dream has come true, with active CDs fully backward compatible while enabling interactive music. The magic is that "the music is in the sound": the structure of the mix is embedded in the sound signal itself, using audio watermarking techniques, and the embedded information is exploited by the player to perform the separation of the sources (patent pending) used in turn by a spatializer.