Classifying and visualizing motion capture sequences using deep neural networks

Kyunghyun Cho, Xi Chen

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

Abstract

The gesture recognition using motion capture data and depth sensors has recently drawn more attention in vision recognition. Currently most systems only classify dataset with a couple of dozens different actions. Moreover, feature extraction from the data is often computational complex. In this paper, we propose a novel system to recognize the actions from skeleton data with simple, but effective, features using deep neural networks. Features are extracted for each frame based on the relative positions of joints (PO), temporal differences (TD), and normalized trajectories of motion (NT). Given these features a hybrid multi-layer perceptron is trained, which simultaneously classifies and reconstructs input data. We use deep autoencoder to visualize learnt features. The experiments show that deep neural networks can capture more discriminative information than, for instance, principal component analysis can. We test our system on a public database with 65 classes and more than 2,000 motion sequences. We obtain an accuracy above 95% which is, to our knowledge, the state of the art result for such a large dataset.

Original languageEnglish (US)
Title of host publicationVISAPP 2014 - Proceedings of the 9th International Conference on Computer Vision Theory and Applications
PublisherSciTePress
Pages122-130
Number of pages9
Volume2
ISBN (Print)9789897580048
StatePublished - 2014
Event9th International Conference on Computer Vision Theory and Applications, VISAPP 2014 - Lisbon, Portugal
Duration: Jan 5 2014Jan 8 2014

Other

Other9th International Conference on Computer Vision Theory and Applications, VISAPP 2014
CountryPortugal
CityLisbon
Period1/5/141/8/14

Fingerprint

Gesture recognition
Multilayer neural networks
Principal component analysis
Feature extraction
Data acquisition
Trajectories
Sensors
Experiments
Deep neural networks

Keywords

  • Deep Neural Network
  • Gesture Recognition
  • Motion Capture

ASJC Scopus subject areas

  • Computer Vision and Pattern Recognition
  • Computer Science Applications

Cite this

Cho, K., & Chen, X. (2014). Classifying and visualizing motion capture sequences using deep neural networks. In VISAPP 2014 - Proceedings of the 9th International Conference on Computer Vision Theory and Applications (Vol. 2, pp. 122-130). SciTePress.

Classifying and visualizing motion capture sequences using deep neural networks. / Cho, Kyunghyun; Chen, Xi.

VISAPP 2014 - Proceedings of the 9th International Conference on Computer Vision Theory and Applications. Vol. 2 SciTePress, 2014. p. 122-130.

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

Cho, K & Chen, X 2014, Classifying and visualizing motion capture sequences using deep neural networks. in VISAPP 2014 - Proceedings of the 9th International Conference on Computer Vision Theory and Applications. vol. 2, SciTePress, pp. 122-130, 9th International Conference on Computer Vision Theory and Applications, VISAPP 2014, Lisbon, Portugal, 1/5/14.
Cho K, Chen X. Classifying and visualizing motion capture sequences using deep neural networks. In VISAPP 2014 - Proceedings of the 9th International Conference on Computer Vision Theory and Applications. Vol. 2. SciTePress. 2014. p. 122-130
Cho, Kyunghyun ; Chen, Xi. / Classifying and visualizing motion capture sequences using deep neural networks. VISAPP 2014 - Proceedings of the 9th International Conference on Computer Vision Theory and Applications. Vol. 2 SciTePress, 2014. pp. 122-130
@inproceedings{b5f3fa93cc4f41068dfc36bcb4fc8211,
title = "Classifying and visualizing motion capture sequences using deep neural networks",
abstract = "The gesture recognition using motion capture data and depth sensors has recently drawn more attention in vision recognition. Currently most systems only classify dataset with a couple of dozens different actions. Moreover, feature extraction from the data is often computational complex. In this paper, we propose a novel system to recognize the actions from skeleton data with simple, but effective, features using deep neural networks. Features are extracted for each frame based on the relative positions of joints (PO), temporal differences (TD), and normalized trajectories of motion (NT). Given these features a hybrid multi-layer perceptron is trained, which simultaneously classifies and reconstructs input data. We use deep autoencoder to visualize learnt features. The experiments show that deep neural networks can capture more discriminative information than, for instance, principal component analysis can. We test our system on a public database with 65 classes and more than 2,000 motion sequences. We obtain an accuracy above 95{\%} which is, to our knowledge, the state of the art result for such a large dataset.",
keywords = "Deep Neural Network, Gesture Recognition, Motion Capture",
author = "Kyunghyun Cho and Xi Chen",
year = "2014",
language = "English (US)",
isbn = "9789897580048",
volume = "2",
pages = "122--130",
booktitle = "VISAPP 2014 - Proceedings of the 9th International Conference on Computer Vision Theory and Applications",
publisher = "SciTePress",

}

TY - GEN

T1 - Classifying and visualizing motion capture sequences using deep neural networks

AU - Cho, Kyunghyun

AU - Chen, Xi

PY - 2014

Y1 - 2014

N2 - The gesture recognition using motion capture data and depth sensors has recently drawn more attention in vision recognition. Currently most systems only classify dataset with a couple of dozens different actions. Moreover, feature extraction from the data is often computational complex. In this paper, we propose a novel system to recognize the actions from skeleton data with simple, but effective, features using deep neural networks. Features are extracted for each frame based on the relative positions of joints (PO), temporal differences (TD), and normalized trajectories of motion (NT). Given these features a hybrid multi-layer perceptron is trained, which simultaneously classifies and reconstructs input data. We use deep autoencoder to visualize learnt features. The experiments show that deep neural networks can capture more discriminative information than, for instance, principal component analysis can. We test our system on a public database with 65 classes and more than 2,000 motion sequences. We obtain an accuracy above 95% which is, to our knowledge, the state of the art result for such a large dataset.

AB - The gesture recognition using motion capture data and depth sensors has recently drawn more attention in vision recognition. Currently most systems only classify dataset with a couple of dozens different actions. Moreover, feature extraction from the data is often computational complex. In this paper, we propose a novel system to recognize the actions from skeleton data with simple, but effective, features using deep neural networks. Features are extracted for each frame based on the relative positions of joints (PO), temporal differences (TD), and normalized trajectories of motion (NT). Given these features a hybrid multi-layer perceptron is trained, which simultaneously classifies and reconstructs input data. We use deep autoencoder to visualize learnt features. The experiments show that deep neural networks can capture more discriminative information than, for instance, principal component analysis can. We test our system on a public database with 65 classes and more than 2,000 motion sequences. We obtain an accuracy above 95% which is, to our knowledge, the state of the art result for such a large dataset.

KW - Deep Neural Network

KW - Gesture Recognition

KW - Motion Capture

UR - http://www.scopus.com/inward/record.url?scp=84906895958&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84906895958&partnerID=8YFLogxK

M3 - Conference contribution

AN - SCOPUS:84906895958

SN - 9789897580048

VL - 2

SP - 122

EP - 130

BT - VISAPP 2014 - Proceedings of the 9th International Conference on Computer Vision Theory and Applications

PB - SciTePress

ER -