Walking Speed Estimation and Gait Classification Using Plantar Pressure and On-Device Deep Learning
Hyuntae Cho
Abstract
Speed estimation and gait classification are widely used in gait analysis for various applications such as medical diagnostics, rehabilitation, and human activity tracking. Smart insoles provide a simple and good infra to measure steps and plantar pressure used for gait analysis. This article proposes a speed estimation and gait classification using plantar pressure and on-device deep learning. First, this article designs and implements a smart insole and Android application to collect plantar pressure data of human beings and then proposes the accurate speed estimation method. Second, I collect various gait datasets from the proposed smart insoles and then build lightweight deep learning models running on the smartphone and tablet PC in real time. In addition, this article evaluated the performance of the proposed speed estimation and ON-device deep learning model. As a result, the accuracy of speed estimation was nearly 97% and that of gait classification was also approximately 97%.