Detecting Disorders in Human Walking Behaviors Using Deep Learning

Abstract

In modern clinics, quantitative analyses are used to diagnose and treat gait pathology. Accordingly, foretasting walking kinematics and kinetics of people benefits to better understand gait patterns and construct assisting devices for rehabilitation. This capstone research proposes a deep learning algorithm (LSTM) for forecasting the walking kinematic of the knee in the Cartesian coordinate system. The hyper-parameter optimization using Gaussian Process method was used to predict walking kinematics of knee with an accuracy of 97%.