|
|
|
Description
|
|
SIMBAD (Smart Inactivity Monitor using Array Based Detectors)
Oct 1999 - May 2001
This project was a collaborative effort involving the University of Manchester, British Telecom,
Infrared Integrated System Ltd. (IRISYS), and the Institute of Human Ageing at the University
of Liverpool. Its aim was to construct a
human fall detector for use in sheltered accommodation units. The system used a low-resolution
infrared sensor produced by IRISYS to monitor a room. When an occupant of the room fell to the ground,
the system would automatically summon medical assistance. British Telecom provided the experience in
telecare systems needed to implement the system commercially. The University of Liverpool provided
sites for field-testing, and information on the dynamics of falls in the elderly. The role of the
University of Manchester was to design algorithms capable of discriminating between falls and normal
movements.
Fractures, particularly of the hip, due to falls at home are a siginificant cause of death in the elderly,
leading to a clear requirement for a system that can summon medical assistance to such incidents.
With the over-60 population of the UK expected to grow by as much as 65% over the coming 30 years, the
world market for this type of device is expected to exceed 10 million units within 7 years.
Current systems used in this application area include CCTV cameras, which are highly intrusive and
requires expensive 24hr monitoring, or personal panic buttons, which rely on the user retaining
concsiousness. The SIMBAD system would be completely autonomous, and would avoid intrusiveness by
using a low-resolution sensor.
The role of the University of Manchester was to design the algorithms required to discriminate falls
from non-fall movements. An actress was recruited to perform a series of falls and non-falls, such as
sitting or lying down, which were recorded
using both the infra-red detector provided by IRISYS and a colour video camera. A colour segmentation
alogrithm was developed to detect the movements of the actress in the colour video sequences, and to
calculate the velocity of her centre of mass. A separate algorithm was developed to determine the
velocity of her centre of mass from the infrared images. The colour video data was then used as a
gold-standard in order to train a neural network to discriminate between the fall and non-fall movements.
The performance of the system was evaluated using a novel implementation of ROC curves. The final system
was then delivered to IRISYS for field trials.
In addition, early work during this project led to the development of the non-parametric image subtraction
algorithm, and to further research into the statistical applications of uniform probabiltiy distributions.
|
|
Contacts
|
|
|
|
Funding
|
|
This project was funded by the MEDLINK program, grant no. P169, supported by the UK Departments of Health, the
Engineering and Physical Sciences Research Council, and the Medical Research Council.
|
|
Software
|
|
This colour segmentation and non-parametric image subtraction algorithms produced during this project are now
available in the TINA machine vision software.
|
|
Documents
|
|
Several Tina Memos describe aspects of this work:
2001-015
Colour Image Segmentation by Non-parametric Density Estimation in Color Space.
2002-004
A Novel Method for Non-Parametric Image Subtraction: Identification of Enhancing Lesions
in Multiple Sclerosis from MR Images.
Publications resulting from this work
Bromiley, P.A., Thacker, N.A., and Courtney, P.
Non-parametric Image Subtraction using Grey Level Scattergrams.
Image and Vision Computing, 20, pp 609-617, 2002.
Bromiley, P.A., Pokric, M., and Thacker, N.A.
Identification of Enhancing MS Lesions in MR Images using Non-Parametric Image Subtraction.
MIUA 2002, Portsmouth, 2002.
Bromiley, P.A., Pokric, M., Thacker, N.A., and Jackson, A.
Detection of MS Lesions in MRI Scans using Non-Parametric Image Subtraction,
ISMRM 2002, Honolulu, Hawaii, 2002.
Bromiley, P.A., Courtney, P., and Thacker, N.A.,
Design of a Visual System for Detecting Natural Events by the use of an Independent Visual
Estimate: a Human Fall Detector.
In Series on Machine Perception and Artificial Intelligence Vol. 50: Empirical Evaluation Methods in Computer Vision.
H.I. Christensen and P.J. Philips (eds.).
World Scientific Publishing, 2002, ISBN 981-02-9453-5.
Bromiley, P.A., Thacker, N.A., and Courtney, P.
Colour Image Segmentation by Non-Parametric Density Estimation in Feature Space.
Proc. BMVC 2001, Manchester, 2001.
Bromiley, P.A., Courtney, P., and Thacker, N.A.
A Case Study in the use of ROC curves for Algorithm Design.
Proc. BMVC 2001, Manchester, 2001.
Bromiley, P.A., Thacker, N.A., and Courtney, P.
Non-parametric Image Subtraction for MRI.
Proc. MIUA 2001, Birmingham, 2001.
Bromiley, P.A., Thacker, N.A., and Courtney, P.
Non-parametric Image Subtraction using Grey-Level Scattergrams.
Proc. British Machine Vision Conference, Bristol, 2000.
|
|
