Detection of Walking Surface Features Using Convolutional Neural Network According to Gait

Abstract

Walking is an integral human activity that includes linked efforts of muscles, brain and nerves. In general, gait of a person is unique and contains useful information belong to that person. Therefore, gait analysis has been used in areas of healthcare, security, sport and fitness. This work differs from other studies by using signals recorded from wearable sensors worn by subjects walking on nine different surfaces. These data and convolutional neural network were employed for walking surface detection. This demonstrates that gait of a person is adopted to walking surface and changes in the gait on different surfaces can be used for walking surface detection. In the experiment, 7958 3-second-long segments were classified with an accuracy of 95%.