Action Recognition Using Motion Primitives and Probabilistic Edit Distance

Research output: Contribution to book/anthology/report/conference proceedingArticle in proceedingResearchpeer-review

12 Citations (Scopus)

Abstract

In this paper we describe a recognition approach based on the notion of primitives. As opposed to recognizing actions based on temporal trajectories or temporal volumes, primitive-based recognition is based on representing a temporal sequence containing an action by only a few characteristic time instances. The human whereabouts at these instances are extracted by double difference images and represented by four features. In each frame the primitive, if any, that best explains the observed data is identified. This leads to a discrete recognition problem since a video sequence will be converted into a string containing a sequence of symbols, each representing a primitives. After pruning the string a probabilistic Edit Distance classifier is applied to identify which action best describes the pruned string. The approach is evaluated on five one-arm gestures and the recognition rate is 91.3%. This is concluded to be a promising result but also leaves room for further improvements.
Original languageEnglish
Title of host publicationArticulated Motion and Deformable Objects
EditorsFrancisco J. Perales, Robert B. Fisher
Number of pages10
PublisherIEEE Computer Society Press
Publication date2006
Pages375-384
ISBN (Print)9783540360315
Publication statusPublished - 2006
Event4th International Conference on Articulated Motion and Deformable Objects - Port d'Andratx, Spain
Duration: 11 Jul 200614 Jul 2006
Conference number: 4

Conference

Conference4th International Conference on Articulated Motion and Deformable Objects
Number4
CountrySpain
CityPort d'Andratx
Period11/07/200614/07/2006
SeriesLecture Notes in Computer Science
Number4069
Volume1
ISSN0302-9743

Fingerprint

Classifiers
Trajectories

Cite this

Fihl, P., Holte, M. B., Moeslund, T. B., & Reng, L. (2006). Action Recognition Using Motion Primitives and Probabilistic Edit Distance. In F. J. Perales, & R. B. Fisher (Eds.), Articulated Motion and Deformable Objects (pp. 375-384). IEEE Computer Society Press. Lecture Notes in Computer Science, No. 4069, Vol.. 1
Fihl, Preben ; Holte, Michael Boelstoft ; Moeslund, Thomas B. ; Reng, Lars. / Action Recognition Using Motion Primitives and Probabilistic Edit Distance. Articulated Motion and Deformable Objects. editor / Francisco J. Perales ; Robert B. Fisher. IEEE Computer Society Press, 2006. pp. 375-384 (Lecture Notes in Computer Science; No. 4069, Vol. 1).
@inproceedings{91917e808a8011dbbb3d000ea68e967b,
title = "Action Recognition Using Motion Primitives and Probabilistic Edit Distance",
abstract = "In this paper we describe a recognition approach based on the notion of primitives. As opposed to recognizing actions based on temporal trajectories or temporal volumes, primitive-based recognition is based on representing a temporal sequence containing an action by only a few characteristic time instances. The human whereabouts at these instances are extracted by double difference images and represented by four features. In each frame the primitive, if any, that best explains the observed data is identified. This leads to a discrete recognition problem since a video sequence will be converted into a string containing a sequence of symbols, each representing a primitives. After pruning the string a probabilistic Edit Distance classifier is applied to identify which action best describes the pruned string. The approach is evaluated on five one-arm gestures and the recognition rate is 91.3{\%}. This is concluded to be a promising result but also leaves room for further improvements.",
author = "Preben Fihl and Holte, {Michael Boelstoft} and Moeslund, {Thomas B.} and Lars Reng",
year = "2006",
language = "English",
isbn = "9783540360315",
series = "Lecture Notes in Computer Science",
publisher = "IEEE Computer Society Press",
number = "4069",
pages = "375--384",
editor = "Perales, {Francisco J.} and Fisher, {Robert B.}",
booktitle = "Articulated Motion and Deformable Objects",
address = "United States",

}

Fihl, P, Holte, MB, Moeslund, TB & Reng, L 2006, Action Recognition Using Motion Primitives and Probabilistic Edit Distance. in FJ Perales & RB Fisher (eds), Articulated Motion and Deformable Objects. IEEE Computer Society Press, Lecture Notes in Computer Science, no. 4069, vol. 1, pp. 375-384, 4th International Conference on Articulated Motion and Deformable Objects, Port d'Andratx, Spain, 11/07/2006.

Action Recognition Using Motion Primitives and Probabilistic Edit Distance. / Fihl, Preben; Holte, Michael Boelstoft; Moeslund, Thomas B.; Reng, Lars.

Articulated Motion and Deformable Objects. ed. / Francisco J. Perales; Robert B. Fisher. IEEE Computer Society Press, 2006. p. 375-384 (Lecture Notes in Computer Science; No. 4069, Vol. 1).

Research output: Contribution to book/anthology/report/conference proceedingArticle in proceedingResearchpeer-review

TY - GEN

T1 - Action Recognition Using Motion Primitives and Probabilistic Edit Distance

AU - Fihl, Preben

AU - Holte, Michael Boelstoft

AU - Moeslund, Thomas B.

AU - Reng, Lars

PY - 2006

Y1 - 2006

N2 - In this paper we describe a recognition approach based on the notion of primitives. As opposed to recognizing actions based on temporal trajectories or temporal volumes, primitive-based recognition is based on representing a temporal sequence containing an action by only a few characteristic time instances. The human whereabouts at these instances are extracted by double difference images and represented by four features. In each frame the primitive, if any, that best explains the observed data is identified. This leads to a discrete recognition problem since a video sequence will be converted into a string containing a sequence of symbols, each representing a primitives. After pruning the string a probabilistic Edit Distance classifier is applied to identify which action best describes the pruned string. The approach is evaluated on five one-arm gestures and the recognition rate is 91.3%. This is concluded to be a promising result but also leaves room for further improvements.

AB - In this paper we describe a recognition approach based on the notion of primitives. As opposed to recognizing actions based on temporal trajectories or temporal volumes, primitive-based recognition is based on representing a temporal sequence containing an action by only a few characteristic time instances. The human whereabouts at these instances are extracted by double difference images and represented by four features. In each frame the primitive, if any, that best explains the observed data is identified. This leads to a discrete recognition problem since a video sequence will be converted into a string containing a sequence of symbols, each representing a primitives. After pruning the string a probabilistic Edit Distance classifier is applied to identify which action best describes the pruned string. The approach is evaluated on five one-arm gestures and the recognition rate is 91.3%. This is concluded to be a promising result but also leaves room for further improvements.

M3 - Article in proceeding

SN - 9783540360315

T3 - Lecture Notes in Computer Science

SP - 375

EP - 384

BT - Articulated Motion and Deformable Objects

A2 - Perales, Francisco J.

A2 - Fisher, Robert B.

PB - IEEE Computer Society Press

ER -

Fihl P, Holte MB, Moeslund TB, Reng L. Action Recognition Using Motion Primitives and Probabilistic Edit Distance. In Perales FJ, Fisher RB, editors, Articulated Motion and Deformable Objects. IEEE Computer Society Press. 2006. p. 375-384. (Lecture Notes in Computer Science; No. 4069, Vol. 1).