A Context-Sensitive Generalization of ICA

Barak A. Pearlmutter, Lucas C. ParraMURAL 1996

Abstract

Source separation arises in a surprising number of signal processing applications, from speech
\nrecognition to EEG analysis. In the square linear blind source separation problem without time delays,
\none must find an unmixing matrix which can detangle the result of mixing n unknown independent sources
\nthrough an unknown n x n mixing matrix. The recently introduced ICA blind source separation algorithm
\n(Baram and Roth 1994; Bell and Sejnowski 1995) is a powerful and surprisingly simple technique for solving
\nthis problem. ICA is all the more remarkable for performing so well despite making absolutely no use of the
\ntemporal structure of its input! This paper presents a new algorithm, contextual ICA, which derives from a
\nmaximum likelihood density estimation formulation of the problem. cICA can incorporate arbitrarily complex
\nadaptive history-sensitive source models, and thereby make use of the temporal structure of its input.
\nThis allows it to separate in a number of situations where standard ICA cannot, including sources with low
\nkurtosis, colored gaussian sources, and sources which have gaussian histograms. Since ICA is a special case
\nof cICA, the MLE derivation provides as a corollary a rigorous derivation of classic ICA.
BibTeX
@inproceedings{pearlmutter1996context,
  author = {Barak A. Pearlmutter and Lucas C. Parra},
  title = {{A Context-Sensitive Generalization of ICA}},
  booktitle = {MURAL},
  year = {1996},
}

Loading related content...

Figure