Human behaviour recognition in data-scarce domains

Rolf Baxter, David Michael Lane, Neil Robertson

Research output: Contribution to journalArticle

Abstract

This paper presents the novel theory for performing multi-agent activity recognition without requiring large training corpora. The reduced need for data means that robust probabilistic recognition can be performed within domains where annotated datasets are traditionally unavailable. Complex human activities are composed from sequences of underlying primitive activities. We do not assume that the exact temporal ordering of primitives is necessary, so can represent complex activity using an unordered bag. Our three-tier architecture comprises low-level video tracking, event analysis and high-level inference. High-level inference is performed using a new, cascading extension of the Rao-Blackwellised Particle Filter. Simulated annealing is used to identify pairs of agents involved in multi-agent activity. We validate our framework using the benchmarked PETS 2006 video surveillance dataset and our own sequences, and achieve a mean recognition F-Score of 0.82. Our approach
achieves a mean improvement of 17% over a Hidden Markov Model baseline.
Original languageEnglish
Pages (from-to)2377-2393
JournalPattern Recognition
Volume48
Issue number8
Early online date6 Mar 2015
DOIs
Publication statusPublished - Aug 2015

Fingerprint

Hidden Markov models
Simulated annealing

Cite this

Baxter, Rolf ; Lane, David Michael ; Robertson, Neil. / Human behaviour recognition in data-scarce domains. In: Pattern Recognition. 2015 ; Vol. 48, No. 8. pp. 2377-2393.
@article{40afe4dc3f9d44c4966cca3e36b16270,
title = "Human behaviour recognition in data-scarce domains",
abstract = "This paper presents the novel theory for performing multi-agent activity recognition without requiring large training corpora. The reduced need for data means that robust probabilistic recognition can be performed within domains where annotated datasets are traditionally unavailable. Complex human activities are composed from sequences of underlying primitive activities. We do not assume that the exact temporal ordering of primitives is necessary, so can represent complex activity using an unordered bag. Our three-tier architecture comprises low-level video tracking, event analysis and high-level inference. High-level inference is performed using a new, cascading extension of the Rao-Blackwellised Particle Filter. Simulated annealing is used to identify pairs of agents involved in multi-agent activity. We validate our framework using the benchmarked PETS 2006 video surveillance dataset and our own sequences, and achieve a mean recognition F-Score of 0.82. Our approachachieves a mean improvement of 17{\%} over a Hidden Markov Model baseline.",
author = "Rolf Baxter and Lane, {David Michael} and Neil Robertson",
year = "2015",
month = "8",
doi = "10.1016/j.patcog.2015.02.019",
language = "English",
volume = "48",
pages = "2377--2393",
journal = "Pattern Recognition",
issn = "0031-3203",
publisher = "Elsevier Limited",
number = "8",

}

Human behaviour recognition in data-scarce domains. / Baxter, Rolf; Lane, David Michael; Robertson, Neil.

In: Pattern Recognition, Vol. 48, No. 8, 08.2015, p. 2377-2393.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Human behaviour recognition in data-scarce domains

AU - Baxter, Rolf

AU - Lane, David Michael

AU - Robertson, Neil

PY - 2015/8

Y1 - 2015/8

N2 - This paper presents the novel theory for performing multi-agent activity recognition without requiring large training corpora. The reduced need for data means that robust probabilistic recognition can be performed within domains where annotated datasets are traditionally unavailable. Complex human activities are composed from sequences of underlying primitive activities. We do not assume that the exact temporal ordering of primitives is necessary, so can represent complex activity using an unordered bag. Our three-tier architecture comprises low-level video tracking, event analysis and high-level inference. High-level inference is performed using a new, cascading extension of the Rao-Blackwellised Particle Filter. Simulated annealing is used to identify pairs of agents involved in multi-agent activity. We validate our framework using the benchmarked PETS 2006 video surveillance dataset and our own sequences, and achieve a mean recognition F-Score of 0.82. Our approachachieves a mean improvement of 17% over a Hidden Markov Model baseline.

AB - This paper presents the novel theory for performing multi-agent activity recognition without requiring large training corpora. The reduced need for data means that robust probabilistic recognition can be performed within domains where annotated datasets are traditionally unavailable. Complex human activities are composed from sequences of underlying primitive activities. We do not assume that the exact temporal ordering of primitives is necessary, so can represent complex activity using an unordered bag. Our three-tier architecture comprises low-level video tracking, event analysis and high-level inference. High-level inference is performed using a new, cascading extension of the Rao-Blackwellised Particle Filter. Simulated annealing is used to identify pairs of agents involved in multi-agent activity. We validate our framework using the benchmarked PETS 2006 video surveillance dataset and our own sequences, and achieve a mean recognition F-Score of 0.82. Our approachachieves a mean improvement of 17% over a Hidden Markov Model baseline.

U2 - 10.1016/j.patcog.2015.02.019

DO - 10.1016/j.patcog.2015.02.019

M3 - Article

VL - 48

SP - 2377

EP - 2393

JO - Pattern Recognition

JF - Pattern Recognition

SN - 0031-3203

IS - 8

ER -