Informed Audio Source Separation from Compressed Linear Stereo Mixtures

In this paper, new developments concerning a system for informed source separation (ISS) of music signals are presented. Such system enables to separate I > 2 musical instruments and singing voices from linear instantaneous stationary stereo (2-channel) mixtures, based on audio signal natural sparsity, pre-mix source signal analysis, and side-information embedding (within the mix signal). The foundations of the system have been presented in previous papers, within the framework of uncompressed (16-bit PCM) mix signals. In the present paper, we extend the study to compressed mix signals. For instance, we use a MPEG-AAC codec and we show that the ISS process is quite robust to compression, opening the way for “real-world” karaoke/soloing/remixing applications for downloadable music.