Learning and generating human natural behaviours for design evaluation using artificial neural networks
Taha, Z.; Brown, R.; Wright, D.
Biomedical Sciences Instrumentation 32: 267-274
1996
ISSN/ISBN: 0067-8856 PMID: 8672678 Document Number: 467899
Biomechanical consideration is becoming very important when designing a product. Animation and strength prediction tools are available to perform the necessary analysis. However with most of these tools, animation is achieved via a sequence of key frames constructed by manipulating the human model to the desired position in each key frame. The resulting motion is therefore unnatural. Most strength predictions are based on static strength measurements of a selected population. Existing prediction tools are not flexible so as to allow data from other populations and/or additional parameters such as dynamic strength to be included in the prediction equations. In this paper we present an approach using neural networks that will allow learning and generation of natural human behaviours. We also propose using neural networks for strength prediction because of the flexibility in specifying inputs and outputs and the ability to map non-linear relationships.