Expected Likelihood sphericity test distribution for complex angular central Gaussian data

Yuri I. Abramovich; Ben A. Johnson; Geoffrey San Antonio

ICASSP 2017accepted0 citations

Expected Likelihood sphericity test distribution for complex angular central Gaussian data

Yuri I. Abramovich, Ben A. Johnson, Geoffrey San Antonio

Abstract

Expected Likelihood based quality assessment of DOA estimates relies on the underlying signal and noise distributions having likelihood ratio probability density functions for the (unknown) true parameters that are independent of the actual true DOAs. This has been shown, both analytically and practically for a wide range of real and complex Gaussian solutions. Recent studies [1, 2] focusing on compound Gaussian mixtures have applied Expected Likelihood based on Monte-Carlo assessment of the “scenario-free” nature of the LR p.d.f.s. In this paper, through specified moments and the use of Mellin's transform, we derive the analytic p.d.f. for the Expected Likelihood sphericity test in the presence of data with the complex angular central Gaussian distribution associated with this compound Gaussian case.

BibTeX

@inproceedings{icassp2017_expectedlikeliho,
  title = {Expected Likelihood sphericity test distribution for complex angular central Gaussian data},
  author = {Yuri I. Abramovich and Ben A. Johnson and Geoffrey San Antonio},
  booktitle = {ICASSP 2017},
  year = {2017}
}