Mean-cyclic-error lower bounds via integral transform of likelihood-ratio function

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

4 Scopus citations

Abstract

In this paper, we consider non-Bayesian periodic parameter estimation and present a new class of mean-cyclic-error (MCE) lower bounds based on integral transform of the likelihood-ratio (LR) function. The MCE bounds in this class are valid for any cyclic-unbiased estimator, in the Lehmann sense, with uniform cyclic performance. Based on the general class of MCE bounds, we propose a novel MCE bound, which utilizes the periodic nature of the problem via the kernel of Fourier series. The proposed bound is based on discrete samples of the LR function in both the frequency and parameter domains and is shown to be tractable and useful for periodic parameter estimation. The proposed bound is compared to the MCE of the maximum-likelihood estimator and to existing MCE bounds in the problem of frequency estimation.

Original languageEnglish
Title of host publication2016 IEEE Sensor Array and Multichannel Signal Processing Workshop, SAM 2016
PublisherIEEE Computer Society
ISBN (Electronic)9781509021031
DOIs
StatePublished - 15 Sep 2016
Event2016 IEEE Sensor Array and Multichannel Signal Processing Workshop, SAM 2016 - Rio de Rio de Janeiro, Brazil
Duration: 10 Jul 201613 Jul 2016

Publication series

NameProceedings of the IEEE Sensor Array and Multichannel Signal Processing Workshop
Volume2016-September
ISSN (Electronic)2151-870X

Conference

Conference2016 IEEE Sensor Array and Multichannel Signal Processing Workshop, SAM 2016
Country/TerritoryBrazil
CityRio de Rio de Janeiro
Period10/07/1613/07/16

Keywords

  • Mean-cyclic-error (MCE) lower bounds
  • cyclic-unbiasedness
  • frequency estimation
  • non-Bayesian periodic parameter estimation
  • uniform cyclic performance

Fingerprint

Dive into the research topics of 'Mean-cyclic-error lower bounds via integral transform of likelihood-ratio function'. Together they form a unique fingerprint.

Cite this