Independent component analysis: Extrema-correspondence properties for higher-order moment functions

Research output: Contribution to conferencePaper

2 Scopus citations

Abstract

Many types of contrast functions have been considered for independent component analysis (ICA). The use of the 4th order kurtosis function and higher-order moment functions for the extraction of ICs from a hybrid mixture are compared. The kurtosis function, based on the 4th order moments is a theoretically attractive contrast function since it possess a traceable extrema-correspondence. Assuming no zero-kurtosis processes are involved, all the only non-negative maxima and non-positive minima of kurtosis function extract at the output. These all and only properties will be referred to as the necessity and sufficiency conditions for the extrema-correspondence.

Original languageEnglish (US)
Pages53-62
Number of pages10
StatePublished - Jan 1 1998
EventProceedings of the 1998 8th IEEE Workshop on Neural Networks for Signal Processing VIII - Cambridge, Engl
Duration: Aug 31 1998Sep 2 1998

Other

OtherProceedings of the 1998 8th IEEE Workshop on Neural Networks for Signal Processing VIII
CityCambridge, Engl
Period8/31/989/2/98

All Science Journal Classification (ASJC) codes

  • Signal Processing
  • Software
  • Electrical and Electronic Engineering

Fingerprint Dive into the research topics of 'Independent component analysis: Extrema-correspondence properties for higher-order moment functions'. Together they form a unique fingerprint.

  • Cite this

    Kung, S. Y. (1998). Independent component analysis: Extrema-correspondence properties for higher-order moment functions. 53-62. Paper presented at Proceedings of the 1998 8th IEEE Workshop on Neural Networks for Signal Processing VIII, Cambridge, Engl, .