Robust GMM Parameter Estimation via the K-BM Algorithm

Ori Kenig, Koby Todros, Tülay Adali

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

1 Scopus citations

Abstract

In this paper, we develop an expectation-maximization (EM)-like scheme, called K-BM, for iterative numerical computation of the minimum K-divergence estimator (MKDE). This estimator utilizes Parzen's non-parameteric Kernel density estimate to down weight low density areas attributed to outliers. Similarly to the standard EM algorithm, the KBM involves successive Maximizations of lower Bounds on the objective function of the MKDE. Differently from EM, these bounds do not rely on conditional expectations only. The proposed K-BM algorithm is applied to robust parameter estimation of a finite-order multivariate Gaussian mixture model (GMM). Simulation studies illustrate the performance advantage of the K-BM as compared to other state-of-the-art robust GMM estimators.

Original languageEnglish
Title of host publicationICASSP 2023 - 2023 IEEE International Conference on Acoustics, Speech and Signal Processing, Proceedings
PublisherInstitute of Electrical and Electronics Engineers
ISBN (Electronic)9781728163277
DOIs
StatePublished - 1 Jan 2023
Event48th IEEE International Conference on Acoustics, Speech and Signal Processing, ICASSP 2023 - Rhodes Island, Greece
Duration: 4 Jun 202310 Jun 2023

Publication series

NameICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings
Volume2023-June
ISSN (Print)1520-6149

Conference

Conference48th IEEE International Conference on Acoustics, Speech and Signal Processing, ICASSP 2023
Country/TerritoryGreece
CityRhodes Island
Period4/06/2310/06/23

Keywords

  • Divergences
  • estimation theory
  • robust statistics

ASJC Scopus subject areas

  • Software
  • Signal Processing
  • Electrical and Electronic Engineering

Fingerprint

Dive into the research topics of 'Robust GMM Parameter Estimation via the K-BM Algorithm'. Together they form a unique fingerprint.

Cite this