Blind Equalization Using the IRWLS Formulation of the Support Vector Machine Articles uri icon

publication date

  • July 2009

start page

  • 1436

end page

  • 1445

issue

  • 7

volume

  • 89

international standard serial number (ISSN)

  • 0165-1684

electronic international standard serial number (EISSN)

  • 1872-7557

abstract

  • In this paper, using a common framework, we propose, analyze, and evaluate several variants of batch algorithms for blind equalization of SISO channels. They are based on the iterative re-weighted least square (IRWLS) solution for the support vector machine (SVM). The proposed methods combine the conventional cost function of the SVM with classical error functions applied to blind equalization: Sato's and Godard's error functions are included in the penalty term of the SVM. The relationship of these batch algorithms with conventional equalization and regularization techniques is analyzed in the paper. Simulation experiments performed over a relevant set of channels show that the proposed equalization methods perform better than traditional cumulant-based methods: they require a lower number of data samples to achieve the same equalization level and convergence ratio.