Depth features to recognise dyadic interactions
Usage of depth sensors in activity recognition is an emerging technology in human–computer interaction. This study presents an approach to recognise human‐to‐human interactions by using depth information. Both hand‐crafted features and deep features extracted from depth frames are studied. After sel...
Main Authors: | , , |
---|---|
Format: | Article |
Language: | English |
Published: |
Wiley
2018-04-01
|
Series: | IET Computer Vision |
Subjects: | |
Online Access: | https://doi.org/10.1049/iet-cvi.2017.0204 |
_version_ | 1797684932413227008 |
---|---|
author | Ali Seydi Keçeli Aydın Kaya Ahmet Burak Can |
author_facet | Ali Seydi Keçeli Aydın Kaya Ahmet Burak Can |
author_sort | Ali Seydi Keçeli |
collection | DOAJ |
description | Usage of depth sensors in activity recognition is an emerging technology in human–computer interaction. This study presents an approach to recognise human‐to‐human interactions by using depth information. Both hand‐crafted features and deep features extracted from depth frames are studied. After selecting and ranking strong features with Relieff algorithm, depth frames are assigned to words. Then, interaction sequences are represented as histograms of words and non‐linear input mapping is applied over histogram bins to minimise differences among various subjects. Random forest, K‐nearest neighbour, and support vector machine (SVM) classifiers are trained using these histograms. The final model is tested on SBU and K3HI datasets and compared with the methods in the literature. In the experiments, joint distances, joint angles and spherical coordinates of the joints were the best performing features. The most successful results are obtained with the composite kernel SVM with Relieff and input mapping methods. While Relieff algorithm helps to select and rank the best features in the feature set, input mapping reduces differences among interactions of various actors. |
first_indexed | 2024-03-12T00:36:57Z |
format | Article |
id | doaj.art-2e96c5f1ea5f4501bb639b19b39dc607 |
institution | Directory Open Access Journal |
issn | 1751-9632 1751-9640 |
language | English |
last_indexed | 2024-03-12T00:36:57Z |
publishDate | 2018-04-01 |
publisher | Wiley |
record_format | Article |
series | IET Computer Vision |
spelling | doaj.art-2e96c5f1ea5f4501bb639b19b39dc6072023-09-15T09:32:17ZengWileyIET Computer Vision1751-96321751-96402018-04-0112333133910.1049/iet-cvi.2017.0204Depth features to recognise dyadic interactionsAli Seydi Keçeli0Aydın Kaya1Ahmet Burak Can2Computer Engineering DepartmentHacettepe University06800AnkaraTurkeyComputer Engineering DepartmentHacettepe University06800AnkaraTurkeyComputer Engineering DepartmentHacettepe University06800AnkaraTurkeyUsage of depth sensors in activity recognition is an emerging technology in human–computer interaction. This study presents an approach to recognise human‐to‐human interactions by using depth information. Both hand‐crafted features and deep features extracted from depth frames are studied. After selecting and ranking strong features with Relieff algorithm, depth frames are assigned to words. Then, interaction sequences are represented as histograms of words and non‐linear input mapping is applied over histogram bins to minimise differences among various subjects. Random forest, K‐nearest neighbour, and support vector machine (SVM) classifiers are trained using these histograms. The final model is tested on SBU and K3HI datasets and compared with the methods in the literature. In the experiments, joint distances, joint angles and spherical coordinates of the joints were the best performing features. The most successful results are obtained with the composite kernel SVM with Relieff and input mapping methods. While Relieff algorithm helps to select and rank the best features in the feature set, input mapping reduces differences among interactions of various actors.https://doi.org/10.1049/iet-cvi.2017.0204depth feature extractiondyadic interaction recognitiondepth sensorsactivity recognitionhuman-computer interactionhuman-to-human interactions |
spellingShingle | Ali Seydi Keçeli Aydın Kaya Ahmet Burak Can Depth features to recognise dyadic interactions IET Computer Vision depth feature extraction dyadic interaction recognition depth sensors activity recognition human-computer interaction human-to-human interactions |
title | Depth features to recognise dyadic interactions |
title_full | Depth features to recognise dyadic interactions |
title_fullStr | Depth features to recognise dyadic interactions |
title_full_unstemmed | Depth features to recognise dyadic interactions |
title_short | Depth features to recognise dyadic interactions |
title_sort | depth features to recognise dyadic interactions |
topic | depth feature extraction dyadic interaction recognition depth sensors activity recognition human-computer interaction human-to-human interactions |
url | https://doi.org/10.1049/iet-cvi.2017.0204 |
work_keys_str_mv | AT aliseydikeceli depthfeaturestorecognisedyadicinteractions AT aydınkaya depthfeaturestorecognisedyadicinteractions AT ahmetburakcan depthfeaturestorecognisedyadicinteractions |