Title: Recognising and predicting gait cycle states for weight-reducing exoskeleton robots using deep learning
Authors: Hanqing Zhao; Hidetaka Nambo
Addresses: Artificial Intelligence Laboratory, Graduate School of Natural Science and Technology, Electrical Engineering and Computer Science, Kanazawa University, Kanazawa, Japan ' Artificial Intelligence Laboratory, Graduate School of Natural Science and Technology, Electrical Engineering and Computer Science, Kanazawa University, Kanazawa, Japan
Abstract: In this study, we propose to use plantar pressure distribution data. Using a combined approach of deep learning and ensemble learning. A proposal for implementing weight-reducing exoskeleton robot in walking gait recognition and temporal prediction. The challenge points of this problem is system architecture design, transformation of data features and morphology, plantar pressure data acquisition and gait data recognition and prediction in real-time. In this paper, we design the system design and implementation for real-time plantar pressure data collection using IoT approach, and real-time gait recognition and prediction using CNN + RNN + ensemble learning model. We designed a plantar pressure measurement device and obtained walking gait datasets through the device. The model is trained using the dataset to obtain a gait recognition and prediction model. Our proposed system solution was implemented for both non-actual walking and actual walking experiments. It is shown experimentally that the gait recognition and gait prediction results of the integrated learning approach in the non-actual walking experiments are 96% and 37%. The gait recognition and gait prediction results of the integrated learning approach in the actual walking experiments are 69% and 42% results.
Keywords: robotic exoskeleton systems; deep learning; IoT; man-machine intelligent system; walk assist.
International Journal of Mechatronics and Automation, 2022 Vol.9 No.1, pp.12 - 21
Received: 15 Sep 2021
Accepted: 26 Sep 2021
Published online: 21 Jan 2022 *