実環境における歩行解析のための靴底センサシステム

抄録

Falls among older adults represent a major public health issue in aging societies such as Japan. Quantitative evaluation of gait function is therefore increasingly important for assessing fall risk and for the clinical evaluation of patients with dizziness and balance disorders. Although three-dimensional motion analysis systems using force plates and optical motion capture systems are considered the gold standard for gait analysis, their high cost and laboratory constraints limit their use in clinical and real-world environments. This article introduces a wearable shoe-based sensor system that integrates multiple triaxial force sensors and an inertial measurement unit (IMU) to evaluate gait biomechanics. The system measures local ground reaction forces at four locations on the sole, and foot motion using an IMU mounted on a toe. By combining these data, the system enables simultaneous estimation of the three-dimensional ground reaction forces (GRFs) and spatiotemporal gait parameters during walking. Machine learning models, including multiple linear regression and Gaussian process regression (GPR), were used to estimate the ground reaction forces from the sensor data. The GRFs estimated using GPR showed good agreement with the results of force plate measurements, with percentage root mean square errors (RMSEs) of less than 10%. In addition, the system estimates minimum toe clearance (MTC), an important parameter associated with trip-related falls. Validation against optical motion capture demonstrated a strong correlation (r = 0.870) and an RMSE of 5.6 mm, which is within the clinically acceptable error range. Applications of the system include gait analysis in older adults, assessment of gait characteristics in stroke patients, and large-scale gait measurements in community-dwelling older populations. The proposed wearable system provides a practical approach for gait analysis outside laboratory environments and may contribute to fall risk evaluation and rehabilitation assessment.