Trajectory-based ambient assistive care for the elderly

Hunter, Andrew and Appiah, Kofi (2010) Trajectory-based ambient assistive care for the elderly. In: RAatE, 29th November 2010, Warwick.

Full text not available from this repository.

Full text URL: http://www.raate.org.uk/2010/rss.xml

Abstract

UN Department of Economic and Social Affairs predicts a worldwide increase in life expectancy coupled with a decline in the number of children and total fertility; thus a drastic reduction in support provided by young individuals working in home and health care for the elderly. This calls for an automated home-based system, capable of monitoring the behaviour and activity level of an elderly living alone independently. The system should have some level of intelligence or decision support properties and yet remain unobtrusive while providing the elderly with an independent life style. This paper presents a smart home system for monitoring the elderly or vulnerable people living independently. The proposed system uses trajectory data extracted from video sensor to infer the level of activity and behaviour.

The proposed video-based home care system presented in this paper has the ability to characterise behaviour as pattern of inactivity, characterise behaviour as pattern of movement and determine a potential fall. We build a probabilistic spatial map of inactivity using the 2D head position of the object and represent it as a mixture of Gaussians (MoG) in 2-dimensional space. The map is used in conjunction with a Hidden Markov Model (HMM) framework, to construct two models: one representing normal activity and the other for an arbitrary behaviour. Similarly the 2D head position is used to model the level of activity and pattern of movement to distinguish between normal and abnormal pattern of movement. By definition, usual or abnormal events are rare, and so lack of training data makes the construction of specific models impractical. Thus unusual behaviour is generally detected by deviation from a model of normal behaviour. This basic assumption is also used to detect fall in our system.

To minimise the amount of data to be transmitted over the network, bearing in mind that video-based home care systems which transmits data is less acceptable, we process the raw video data and only transmit the vital 2D data for further analysis at a central hub. This does not only make the wireless transmission feasible, but also such a system will be more acceptable in a care home than a similar system which transmits raw video data. Again, the proposed system is based on a wide angle “fish-eye lens” camera, making it less obstructive while reducing the installation cost. The system is energy efficient and can be battery powered. Also power hungry video processing is only triggered when there is sufficient movement in the environment, saving approximately 80% of power when monitoring an elderly. The proposed assistive care system uses low cost sensors, can be battery powered and very easy to install in any indoor environment. It has the ability to generate real-time alerts like falls, less than expected or more than expected movements or activity, and unexpected pattern of movement in the house.

Item Type:Conference or Workshop Item (Other)
Additional Information:UN Department of Economic and Social Affairs predicts a worldwide increase in life expectancy coupled with a decline in the number of children and total fertility; thus a drastic reduction in support provided by young individuals working in home and health care for the elderly. This calls for an automated home-based system, capable of monitoring the behaviour and activity level of an elderly living alone independently. The system should have some level of intelligence or decision support properties and yet remain unobtrusive while providing the elderly with an independent life style. This paper presents a smart home system for monitoring the elderly or vulnerable people living independently. The proposed system uses trajectory data extracted from video sensor to infer the level of activity and behaviour. The proposed video-based home care system presented in this paper has the ability to characterise behaviour as pattern of inactivity, characterise behaviour as pattern of movement and determine a potential fall. We build a probabilistic spatial map of inactivity using the 2D head position of the object and represent it as a mixture of Gaussians (MoG) in 2-dimensional space. The map is used in conjunction with a Hidden Markov Model (HMM) framework, to construct two models: one representing normal activity and the other for an arbitrary behaviour. Similarly the 2D head position is used to model the level of activity and pattern of movement to distinguish between normal and abnormal pattern of movement. By definition, usual or abnormal events are rare, and so lack of training data makes the construction of specific models impractical. Thus unusual behaviour is generally detected by deviation from a model of normal behaviour. This basic assumption is also used to detect fall in our system. To minimise the amount of data to be transmitted over the network, bearing in mind that video-based home care systems which transmits data is less acceptable, we process the raw video data and only transmit the vital 2D data for further analysis at a central hub. This does not only make the wireless transmission feasible, but also such a system will be more acceptable in a care home than a similar system which transmits raw video data. Again, the proposed system is based on a wide angle “fish-eye lens” camera, making it less obstructive while reducing the installation cost. The system is energy efficient and can be battery powered. Also power hungry video processing is only triggered when there is sufficient movement in the environment, saving approximately 80% of power when monitoring an elderly. The proposed assistive care system uses low cost sensors, can be battery powered and very easy to install in any indoor environment. It has the ability to generate real-time alerts like falls, less than expected or more than expected movements or activity, and unexpected pattern of movement in the house.
Keywords:Assistive Care, Trajectory classification, home care
Subjects:G Mathematical and Computer Sciences > G700 Artificial Intelligence
G Mathematical and Computer Sciences > G740 Computer Vision
Divisions:College of Science > School of Computer Science
ID Code:4385
Deposited By:INVALID USER
Deposited On:08 Apr 2011 19:03
Last Modified:18 Jul 2011 16:41

Repository Staff Only: item control page