- Author
- Mohamed Eldib, Francis Deboeverie (UGent) , Dirk Van Haerenborgh (UGent) , Somaya Ben Allouch, Wilfried Philips (UGent) and Hamid Aghajan (UGent)
- Organization
- Abstract
- Loneliness and social isolation are probably the most prevalent psychosocial problems related to aging. One critical component in assessing social isolation in an unobtrusive manner is to measure the out-of-home activity levels, as social isolation often goes along with decreased physical activity, decreased motoric functioning, and a decline in activities of daily living, all of which may lead to a reduction in the amount of time spent out-of-home. In this work, we propose to use a single visual sensor for detecting out-of-home activity. The visual sensor has a very low spatial resolution (900 pixels), which is a key feature to ensure a cheap technology and to maintain the user’s privacy. Firstly, the visual sensor is installed in a top view setup at the door entrance. Secondly, a correlation-based foreground detection method is used to extract the foreground. Thirdly, an Extra Trees Classifier (ETC) is trained to classify the directionality of the person (in/out) based on the motion of the foreground pixels. Due to the nature of variability of the out-of-home activity, the relative frequency of the directionality (in/out) is measured over a window of 3 seconds to determine the final result. We installed our system in 9 different service flats in the UK, Belgium and France where the same ETC model is used. We evaluate our method on video sequences captured in real-life environments from the different setups, where the persons’ out-of-home routines are recorded. The results show that our approach of detecting out-of-home activity achieves an accuracy of 91.30%.
- Keywords
- visual sensor, out-of-home, Ambient assisted living (AAL)
Downloads
-
810b61edab6040bdb8e1c8e5bf8bfd5d.pdf
- full text
- |
- open access
- |
- |
- 169.53 KB
Citation
Please use this url to cite or link to this publication: http://hdl.handle.net/1854/LU-7221910
- MLA
- Eldib, Mohamed, et al. Out-of-Home Activity Analysis Using a Low-Resolution Visual Sensor. Vol. 15, Gerontechnology, 2016, pp. 79–79, doi:10.4017/gt.2016.15.s.825.00.
- APA
- Eldib, M., Deboeverie, F., Van Haerenborgh, D., Ben Allouch, S., Philips, W., & Aghajan, H. (2016). Out-of-home activity analysis using a low-resolution visual sensor. 15, 79–79. https://doi.org/10.4017/gt.2016.15.s.825.00
- Chicago author-date
- Eldib, Mohamed, Francis Deboeverie, Dirk Van Haerenborgh, Somaya Ben Allouch, Wilfried Philips, and Hamid Aghajan. 2016. “Out-of-Home Activity Analysis Using a Low-Resolution Visual Sensor.” In , 15:79–79. Gerontechnology. https://doi.org/10.4017/gt.2016.15.s.825.00.
- Chicago author-date (all authors)
- Eldib, Mohamed, Francis Deboeverie, Dirk Van Haerenborgh, Somaya Ben Allouch, Wilfried Philips, and Hamid Aghajan. 2016. “Out-of-Home Activity Analysis Using a Low-Resolution Visual Sensor.” In , 15:79–79. Gerontechnology. doi:10.4017/gt.2016.15.s.825.00.
- Vancouver
- 1.Eldib M, Deboeverie F, Van Haerenborgh D, Ben Allouch S, Philips W, Aghajan H. Out-of-home activity analysis using a low-resolution visual sensor. In Gerontechnology; 2016. p. 79–79.
- IEEE
- [1]M. Eldib, F. Deboeverie, D. Van Haerenborgh, S. Ben Allouch, W. Philips, and H. Aghajan, “Out-of-home activity analysis using a low-resolution visual sensor,” presented at the 10th World Conference of Gerontechnology (ISG 2016), Nice, France, 2016, vol. 15, pp. 79–79.
@inproceedings{7221910,
abstract = {{Loneliness and social isolation are probably the most prevalent
psychosocial problems related to aging. One critical component in
assessing social isolation in an unobtrusive manner is to measure the
out-of-home activity levels, as social isolation often goes along with
decreased physical activity, decreased motoric functioning, and a decline
in activities of daily living, all of which may lead to a reduction in the
amount of time spent out-of-home. In this work, we propose to use a
single visual sensor for detecting out-of-home activity. The visual
sensor has a very low spatial resolution (900 pixels), which is a key
feature to ensure a cheap technology and to maintain the user’s
privacy. Firstly, the visual sensor is installed in a top view setup at
the door entrance. Secondly, a correlation-based foreground detection
method is used to extract the foreground. Thirdly, an Extra Trees
Classifier (ETC) is trained to classify the directionality of the person
(in/out) based on the motion of the foreground pixels. Due to the nature
of variability of the out-of-home activity, the relative frequency of
the directionality (in/out) is measured over a window of 3 seconds to
determine the final result. We installed our system in 9 different
service flats in the UK, Belgium and France where the same ETC model is
used. We evaluate our method on video sequences captured in real-life
environments from the different setups, where the persons’ out-of-home
routines are recorded. The results show that our approach of detecting
out-of-home activity achieves an accuracy of 91.30%.}},
author = {{Eldib, Mohamed and Deboeverie, Francis and Van Haerenborgh, Dirk and Ben Allouch, Somaya and Philips, Wilfried and Aghajan, Hamid}},
issn = {{1569-1101}},
keywords = {{visual sensor,out-of-home,Ambient assisted living (AAL)}},
language = {{eng}},
location = {{Nice, France}},
pages = {{79--79}},
publisher = {{Gerontechnology}},
title = {{Out-of-home activity analysis using a low-resolution visual sensor}},
url = {{http://doi.org/10.4017/gt.2016.15.s.825.00}},
volume = {{15}},
year = {{2016}},
}
- Altmetric
- View in Altmetric