Human action recognition by SOM considering the probability of spatio-temporal features

Research output: Chapter in Book/Report/Conference proceedingConference contribution

3 Citations (Scopus)

Abstract

In this paper, an action recognition system was invented by proposing a compact 3D descriptor to represent action information, and employing self-organizing map (SOM) to learn and recognize actions. Histogram Of Gradient 3D (HOG3D) performed better among currently used descriptors for action recognition. However, the calculation of the descriptor is quite complex. Furthermore, it used a vector with 960 elements to describe one interest point. Therefore, we proposed a compact descriptor, which shortened the support region of interest points, combined symmetric bins after orientation quantization. In addition, the top value bin of quantized vector was kept instead of setting threshold experimentally. Comparing with HOG3D, our descriptor used 80 bins to describe a point, which reduced much computation complexity. The compact descriptor was used to learn and recognize actions considering the probability of local features in SOM, and the results showed that our system outperformed others both on KTH and Hollywood datasets.

Original languageEnglish
Title of host publicationNeural Information Processing
Subtitle of host publicationModels and Applications - 17th International Conference, ICONIP 2010, Proceedings
Pages391-398
Number of pages8
EditionPART 2
DOIs
Publication statusPublished - Dec 21 2010
Event17th International Conference on Neural Information Processing, ICONIP 2010 - Sydney, NSW, Australia
Duration: Nov 22 2010Nov 25 2010

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
NumberPART 2
Volume6444 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Other

Other17th International Conference on Neural Information Processing, ICONIP 2010
CountryAustralia
CitySydney, NSW
Period11/22/1011/25/10

Fingerprint

Action Recognition
Self organizing maps
Bins
Self-organizing Map
Descriptors
Histogram
Gradient
Local Features
Region of Interest
Human
Quantization

All Science Journal Classification (ASJC) codes

  • Theoretical Computer Science
  • Computer Science(all)

Cite this

Ji, Y., Shimada, A., & Taniguchi, R-I. (2010). Human action recognition by SOM considering the probability of spatio-temporal features. In Neural Information Processing: Models and Applications - 17th International Conference, ICONIP 2010, Proceedings (PART 2 ed., pp. 391-398). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 6444 LNCS, No. PART 2). https://doi.org/10.1007/978-3-642-17534-3_48

Human action recognition by SOM considering the probability of spatio-temporal features. / Ji, Yanli; Shimada, Atsushi; Taniguchi, Rin-Ichiro.

Neural Information Processing: Models and Applications - 17th International Conference, ICONIP 2010, Proceedings. PART 2. ed. 2010. p. 391-398 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 6444 LNCS, No. PART 2).

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Ji, Y, Shimada, A & Taniguchi, R-I 2010, Human action recognition by SOM considering the probability of spatio-temporal features. in Neural Information Processing: Models and Applications - 17th International Conference, ICONIP 2010, Proceedings. PART 2 edn, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), no. PART 2, vol. 6444 LNCS, pp. 391-398, 17th International Conference on Neural Information Processing, ICONIP 2010, Sydney, NSW, Australia, 11/22/10. https://doi.org/10.1007/978-3-642-17534-3_48
Ji Y, Shimada A, Taniguchi R-I. Human action recognition by SOM considering the probability of spatio-temporal features. In Neural Information Processing: Models and Applications - 17th International Conference, ICONIP 2010, Proceedings. PART 2 ed. 2010. p. 391-398. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); PART 2). https://doi.org/10.1007/978-3-642-17534-3_48
Ji, Yanli ; Shimada, Atsushi ; Taniguchi, Rin-Ichiro. / Human action recognition by SOM considering the probability of spatio-temporal features. Neural Information Processing: Models and Applications - 17th International Conference, ICONIP 2010, Proceedings. PART 2. ed. 2010. pp. 391-398 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); PART 2).
@inproceedings{49cdb9c489cf434e879d688dd5ce4d9d,
title = "Human action recognition by SOM considering the probability of spatio-temporal features",
abstract = "In this paper, an action recognition system was invented by proposing a compact 3D descriptor to represent action information, and employing self-organizing map (SOM) to learn and recognize actions. Histogram Of Gradient 3D (HOG3D) performed better among currently used descriptors for action recognition. However, the calculation of the descriptor is quite complex. Furthermore, it used a vector with 960 elements to describe one interest point. Therefore, we proposed a compact descriptor, which shortened the support region of interest points, combined symmetric bins after orientation quantization. In addition, the top value bin of quantized vector was kept instead of setting threshold experimentally. Comparing with HOG3D, our descriptor used 80 bins to describe a point, which reduced much computation complexity. The compact descriptor was used to learn and recognize actions considering the probability of local features in SOM, and the results showed that our system outperformed others both on KTH and Hollywood datasets.",
author = "Yanli Ji and Atsushi Shimada and Rin-Ichiro Taniguchi",
year = "2010",
month = "12",
day = "21",
doi = "10.1007/978-3-642-17534-3_48",
language = "English",
isbn = "3642175333",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
number = "PART 2",
pages = "391--398",
booktitle = "Neural Information Processing",
edition = "PART 2",

}

TY - GEN

T1 - Human action recognition by SOM considering the probability of spatio-temporal features

AU - Ji, Yanli

AU - Shimada, Atsushi

AU - Taniguchi, Rin-Ichiro

PY - 2010/12/21

Y1 - 2010/12/21

N2 - In this paper, an action recognition system was invented by proposing a compact 3D descriptor to represent action information, and employing self-organizing map (SOM) to learn and recognize actions. Histogram Of Gradient 3D (HOG3D) performed better among currently used descriptors for action recognition. However, the calculation of the descriptor is quite complex. Furthermore, it used a vector with 960 elements to describe one interest point. Therefore, we proposed a compact descriptor, which shortened the support region of interest points, combined symmetric bins after orientation quantization. In addition, the top value bin of quantized vector was kept instead of setting threshold experimentally. Comparing with HOG3D, our descriptor used 80 bins to describe a point, which reduced much computation complexity. The compact descriptor was used to learn and recognize actions considering the probability of local features in SOM, and the results showed that our system outperformed others both on KTH and Hollywood datasets.

AB - In this paper, an action recognition system was invented by proposing a compact 3D descriptor to represent action information, and employing self-organizing map (SOM) to learn and recognize actions. Histogram Of Gradient 3D (HOG3D) performed better among currently used descriptors for action recognition. However, the calculation of the descriptor is quite complex. Furthermore, it used a vector with 960 elements to describe one interest point. Therefore, we proposed a compact descriptor, which shortened the support region of interest points, combined symmetric bins after orientation quantization. In addition, the top value bin of quantized vector was kept instead of setting threshold experimentally. Comparing with HOG3D, our descriptor used 80 bins to describe a point, which reduced much computation complexity. The compact descriptor was used to learn and recognize actions considering the probability of local features in SOM, and the results showed that our system outperformed others both on KTH and Hollywood datasets.

UR - http://www.scopus.com/inward/record.url?scp=78650192992&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=78650192992&partnerID=8YFLogxK

U2 - 10.1007/978-3-642-17534-3_48

DO - 10.1007/978-3-642-17534-3_48

M3 - Conference contribution

AN - SCOPUS:78650192992

SN - 3642175333

SN - 9783642175336

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 391

EP - 398

BT - Neural Information Processing

ER -