Wireless Smart Shoes for Running Gait Analysis Based on Deep Learning

Abstract

In this work, we propose a design of wireless smart shoes for running gait analysis. The system is based on an inertial measurement unit (IMU) and Bluetooth low energy (BLE) protocol to save energy. The estimated parameters include activity type (run/walk/rest), forward distance, and average moving speed. These parameters are calculated and displayed on the user’s smartphone. Different from previous works, IMU data is collected to train a compact neural network model for running gait classification. The performance of model prediction and other measurements is evaluated with a customized database of 600 data segments. Experimental results show that our model achieves a high accuracy of 99.35% on gait classification. Other measures for moving analysis such as the total forward distance and the average speed are retrieved with a low maximum error of less than 4.67% and 4.80% respectively.