Detecting Perturbation Occurrence during Walking

This research aims at the development and verification of a system that can detect the occurrence of perturbation during walking for walking-assist as soon as possible by from output signals of an artificial sensor system fitted on users’ body. In this research, a back-propagation based artificial neural network (ANN) model was employed to build pattern recognition unit that can deal with the individual variation and time-varying characteristics. Results showed that the occurrence could be detected from four channels within 87.5ms, which could be considered as short enough for walking assist systems to response to the perturbation.