Development of Shokac-Shoes using MEMS 6-axis force-torque sensors

Abstract

We developed "Shokac-shoes" by placing three MEMS 6-axis force-torque sensors and a 6-axis inertial sensor on the insole of each shoe. Sensor data acquired by the “Shokac-shoes” were transferred to the smartphone via Bluetooth and then uploaded to the cloud server. Two AI analysis models were introduced to estimate the ground reaction force (GRF) and foot lift height from the limited data of 24 sensors. One was constructed using deep neural network for estimating GRF, and another was a combination of convolutional neural network and bidirectional LSTM (Long Short-Term Memory) for estimating foot lift height. It was confirmed that GRF and foot lift height could be estimated with validity by inputting time-series sensor data into the constructed AI models. The time required for AI models to present the estimated result to the user was about 5 seconds, which was confirmed to be short enough for practical use as the waiting time in the field of gait analysis. It is expected that the reduction of data costs by using “Shokac-shoes” will further promote research in various fields such as sports, healthcare and medicine.