The purpose of this study was to estimate the spatiotemporal gait parameters from step time information during walking-in-place (WIP) and body anthropometric information from a newly developed VR locomotion system using a learning-based regressor. A fully-connected feed-forward neural network model was used to predict the spatiotemporal variables of walking. The inputs of the model were the WIP features and body anthropometric data, while the outputs of the model were the spatiotemporal gait parameters of the regular walking. With the prediction accuracy of 98% or higher, the feasibility of the model has been validated. In conclusion, the model not only can provide accurate prediction of spatiotemporal gait parameters while the users are walking in the VR locomotion system, but also eliminate the need to measure these parameters in experimental environments. Future studies with various subject groups such as the elderly and patients with musculoskeletal injuries will be conducted to generalize the findings of this study.
Mun, Kyung-Ryoul; Jeong, Hwansu; Oh, Hyungan; and Kim, Jinwook
"A LEARNING-BASED GAIT ESTIMATION DURING WALKING-IN-PLACE IN VR LOCOMOTION SYSTEM,"
ISBS Proceedings Archive: Vol. 36:
1, Article 140.
Available at: https://commons.nmu.edu/isbs/vol36/iss1/140