Non-Gaussian multivariate adaptive AR estimation using the superexponential algorithm

  • Authors:

  • M. Martone

  • Affiliations:

  • Telecomm. Group, Watkins-Johnson Co., Gaithersburg, MD

  • Venue:
  • IEEE Transactions on Signal Processing
  • Year:
  • 1996

Quantified Score

Hi-index 35.68

Visualization

Abstract

We formulate as a deconvolution problem the causal/noncausal non-Gaussian multichannel autoregressive (AR) parameter estimation problem. The super exponential algorithm presented in a paper by Shalvi and Weinstein (1993) is generalized to the vector case. We present an adaptive implementation that is very attractive since it is higher order statistics (HOS) based but does not present the high computational complexity of methods proposed up to now