Audio Source Separation
Mike Davies · 2002
Abstract In 1991 Jutten [10] proposed a Neural Network for separating out independent signals from a set of mixtures. This has led to the development of Independent Component Analysis (ICA) or Blind Signal Separation (BSS). The problem is deceivingly simple. Currently a number of extensions to ICA are being explored. Of particular interest in audio is the extension to convolutional mixing since an acoustic environment typically includes delays, echoes and reverberation. However, we are by no means the first to tackle the convolutional mixing problem and a number of algorithms have been proposed [5, 16, 17, 11, 13, 15].