Classification of Hemiparetic Gait and Normal Gait according to Soleus EMG Signal using Deep Learning Method

Abstract

Soleus EMG signals were collected from normal subjects and chronic stroke patients during overground gait. The signals from each gait cycle were converted into continuous Wavelet transform (CWT) images. The training images were used to train 2 types of Convolutional Neural Network (CNN). The trained networks classified the test images into normal gait and hemiparetic gait with over 98% accuracy.