Motion Intention Estimation Using Machine Learning with Surface ElectroMyoGraphy

-Comparison of Accuracy between Support Vector Machine and Random Forest-

Abstract

Accidents involving power wheelchairs occur frequently among the elderly, and one of the causes is a decline in cognitive reaction time due to aging. We aim to reduce power wheelchair accidents by improving the cognitive reaction time of the elderly. For this purpose, we measured surface electromyography (sEMG), a type of biological signal that appears before movement. The subject is asked to perform an action that simulates a power wheelchair, and the surface sEMG is measured. The measured sEMGs are computed and analyzed as features for machine learning to estimate the distance and direction of the subject's hand movement. Two types of machine learning are used: SVM and Random Forest. We analyze the results using these machine learning methods and compare their accuracy.