Efficient particle filtering via sparse kernel density estimation

Amit Banerjee, Philippe Burlina

Research output: Contribution to journalArticle

Abstract

Particle filters (PFs) are Bayesian filters capable of modeling nonlinear, non-Gaussian, and nonstationary dynamical systems. Recent research in PFs has investigated ways to appropriately sample from the posterior distribution, maintain multiple hypotheses, and alleviate computational costs while preserving tracking accuracy. To address these issues, a novel utilization of the support vector data description (SVDD) density estimation method within the particle filtering framework is presented. The SVDD density estimate can be integrated into a wide range of PFs to realize several benefits. It yields a sparse representation of the posterior density that reduces the computational complexity of the PF. The proposed approach also provides an analytical expression for the posterior distribution that can be used to identify its modes for maintaining multiple hypotheses and computing the MAP estimate, and to directly sample from the posterior. We present several experiments that demonstrate the advantages of incorporating a sparse kernel density estimate in a particle filter.

Original languageEnglish (US)
Article number5443441
Pages (from-to)2480-2490
Number of pages11
JournalIEEE Transactions on Image Processing
Volume19
Issue number9
DOIs
StatePublished - Sep 1 2010

Fingerprint

Data description
Computational complexity
Dynamical systems
Costs
Experiments

Keywords

  • Bayesian filtering
  • machine learning
  • particle filters
  • support vectors
  • tracking

ASJC Scopus subject areas

  • Software
  • Computer Graphics and Computer-Aided Design

Cite this

Efficient particle filtering via sparse kernel density estimation. / Banerjee, Amit; Burlina, Philippe.

In: IEEE Transactions on Image Processing, Vol. 19, No. 9, 5443441, 01.09.2010, p. 2480-2490.

Research output: Contribution to journalArticle

Banerjee, Amit ; Burlina, Philippe. / Efficient particle filtering via sparse kernel density estimation. In: IEEE Transactions on Image Processing. 2010 ; Vol. 19, No. 9. pp. 2480-2490.
@article{e1cba01881904a8f9c7010c031a12da9,
title = "Efficient particle filtering via sparse kernel density estimation",
abstract = "Particle filters (PFs) are Bayesian filters capable of modeling nonlinear, non-Gaussian, and nonstationary dynamical systems. Recent research in PFs has investigated ways to appropriately sample from the posterior distribution, maintain multiple hypotheses, and alleviate computational costs while preserving tracking accuracy. To address these issues, a novel utilization of the support vector data description (SVDD) density estimation method within the particle filtering framework is presented. The SVDD density estimate can be integrated into a wide range of PFs to realize several benefits. It yields a sparse representation of the posterior density that reduces the computational complexity of the PF. The proposed approach also provides an analytical expression for the posterior distribution that can be used to identify its modes for maintaining multiple hypotheses and computing the MAP estimate, and to directly sample from the posterior. We present several experiments that demonstrate the advantages of incorporating a sparse kernel density estimate in a particle filter.",
keywords = "Bayesian filtering, machine learning, particle filters, support vectors, tracking",
author = "Amit Banerjee and Philippe Burlina",
year = "2010",
month = "9",
day = "1",
doi = "10.1109/TIP.2010.2047667",
language = "English (US)",
volume = "19",
pages = "2480--2490",
journal = "IEEE Transactions on Image Processing",
issn = "1057-7149",
publisher = "Institute of Electrical and Electronics Engineers Inc.",
number = "9",

}

TY - JOUR

T1 - Efficient particle filtering via sparse kernel density estimation

AU - Banerjee, Amit

AU - Burlina, Philippe

PY - 2010/9/1

Y1 - 2010/9/1

N2 - Particle filters (PFs) are Bayesian filters capable of modeling nonlinear, non-Gaussian, and nonstationary dynamical systems. Recent research in PFs has investigated ways to appropriately sample from the posterior distribution, maintain multiple hypotheses, and alleviate computational costs while preserving tracking accuracy. To address these issues, a novel utilization of the support vector data description (SVDD) density estimation method within the particle filtering framework is presented. The SVDD density estimate can be integrated into a wide range of PFs to realize several benefits. It yields a sparse representation of the posterior density that reduces the computational complexity of the PF. The proposed approach also provides an analytical expression for the posterior distribution that can be used to identify its modes for maintaining multiple hypotheses and computing the MAP estimate, and to directly sample from the posterior. We present several experiments that demonstrate the advantages of incorporating a sparse kernel density estimate in a particle filter.

AB - Particle filters (PFs) are Bayesian filters capable of modeling nonlinear, non-Gaussian, and nonstationary dynamical systems. Recent research in PFs has investigated ways to appropriately sample from the posterior distribution, maintain multiple hypotheses, and alleviate computational costs while preserving tracking accuracy. To address these issues, a novel utilization of the support vector data description (SVDD) density estimation method within the particle filtering framework is presented. The SVDD density estimate can be integrated into a wide range of PFs to realize several benefits. It yields a sparse representation of the posterior density that reduces the computational complexity of the PF. The proposed approach also provides an analytical expression for the posterior distribution that can be used to identify its modes for maintaining multiple hypotheses and computing the MAP estimate, and to directly sample from the posterior. We present several experiments that demonstrate the advantages of incorporating a sparse kernel density estimate in a particle filter.

KW - Bayesian filtering

KW - machine learning

KW - particle filters

KW - support vectors

KW - tracking

UR - http://www.scopus.com/inward/record.url?scp=77955828051&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=77955828051&partnerID=8YFLogxK

U2 - 10.1109/TIP.2010.2047667

DO - 10.1109/TIP.2010.2047667

M3 - Article

C2 - 20371403

AN - SCOPUS:77955828051

VL - 19

SP - 2480

EP - 2490

JO - IEEE Transactions on Image Processing

JF - IEEE Transactions on Image Processing

SN - 1057-7149

IS - 9

M1 - 5443441

ER -