Predicting Walking Intentions using sEMG and Mechanical sensors for various environment

Predicting the motion intentions of a user is very challenging when controlling an exoskeleton robot. When only a mechanical sensor is used, a change in the motion is detected during the user's movement. An electromyographic (EMG) signal, which is a biological signal, is detected by the activation of the muscles before the actual movement of a person. Using the EMG signal, the motion intention can be identified before the actual movement, and the delay in time in controlling the exoskeletal robot can be shortened to reduce the resistance felt by the user. In this paper, the surface electromyographic (sEMG) signal is used together with a mechanical sensor to identify the walking environment according to the walking gait cycle. In the classification, the combination of sensors was varied, and information from one leg and two legs was analyzed by the different gait periods before and after heel contact and toe off. As a result of the classification into three sensor combinations, sEMG, kinetic, and kinematic sensors, at the pre heel contact time before walking, a 96.8% and 98.6% accuracy was obtained for information from one and two legs, respectively. In the same gait environment, it was shown that the gait prediction can be performed based on the time unit by dividing the time interval before starting the gait. An average accuracy of 84.4% was obtained when the time was divided by the environment in intervals of 100ms before heel contact, and the average was 90.9% when it was divided by an interval of 200ms before heel contact.