Novel Walking-Intention Recognition Method for Omnidirectional Walking Support Robot

Abstract

Walking support robots have attracted much attention from the public. In previous research, the walking support robot (WSR) system had the omnidirectional movement function and used forearm pressure to control the WSR. While this paper mainly discusses how to improve recognition accuracy of the user’s walking directional intention. A novel walking-intention recognition method is proposed which fuses and computes the data of acceleration and forearm pressure to identify the user’s walking directional intention. To be specific, firstly, a smartphone is fixed to the user's chest to obtain the data from accelerometer of the smartphone, and the forearm pressure imposed on the WSR by the users’ wrists and elbows are measured by six force sensors embedded in the WSR’s armrest. Secondly, fusing and computing the data of these sensors can be complementary to attain improved recognition accuracy and a better efficiency. Thirdly, a support vector machine algorithm (SVM) is presented to estimate the user’s walk directional intention. Finally, the validity of the proposed directional identification method is experimentally confirmed.