Skip directly to search Skip directly to A to Z list Skip directly to page options Skip directly to site content

NIOSHTIC-2 Publications Search

Search Results

A back-propagation neural network model for prediction of loss of balance.

Wang-W; Bhattacharya-A
Proceedings of the 1996 Fifteenth Southern Biomedical Engineering Conference, Dayton, Ohio, March 29-31, 1996. Bajpai PK, ed., Piscataway, NJ: Institute of Electrical Engineers, 1996 Mar; :85-88
Neural network models for the prediction of the postural sway response to occupational risk factors including environmental lighting, job tasks, floor surface, work load, peripheral vision, age, and gender were presented. Two five layer back propagation networks were developed, trained, and tested using the Professional Neural Network Developer. Two variables of sway, Index of Proximity to Stability Boundary (IPSB), measuring the minimum distance between the body's center of pressure (CP) and the boundary of supporting base area, and sway length (SL), measuring the distance traveled by the body's CP during the test period, were used. Risk factors included workload level, compliant or firm surfaces, dry or oil surfaces, dim or bright lighting, unobstructed or obstructed view, stationary, bending, upward reach and sudden loading tasks, various age groups, and gender. Values predicted from the neural network models after 50,000 iterations were very close to expected values. Data from a project examining the effect of fall risk factors on the ability of workers to maintain an upright balance conducted in 52 male and female workers under 10 test conditions were used as the training set. Predictions using these data applied as inputs to the trained networks were also close to expected values demonstrating the ability of the models to make valid predictions in unknown situations. The effect of each risk factor on the output variables was examined using the trained network models. Such analyses suggested that age was a risk factor with a direct impact on IPSB and SL, and that changes in task type increased the level of balance loss. The authors conclude that artificial neural network models are useful in the determination of risk factors affecting loss of balance and in the development of intervention programs.
NIOSH-Grant; Traumatic-injuries; Computer-models; Mathematical-models; Analytical-methods; Biomechanics; Ergonomics; Posture; Age-factors; Task-performance; Work-analysis
Environmental Health University of Cincinnati 3223 Eden Ave Cincinnati, OH 45267-0056
Publication Date
Document Type
Conference/Symposia Proceedings
Funding Amount
Funding Type
Fiscal Year
NTIS Accession No.
NTIS Price
Identifying No.
Source Name
Proceedings of the 1996 Fifteenth Southern Biomedical Engineering Conference, Dayton, Ohio, March 29-31, 1996
Performing Organization
University of Cincinnati