The Lower Limbs Kinematics Analysis by Wearable Sensor Shoes

This is the authors' abstract. We don't add key points for this paper.

Inertial sensors designed for wearable devices in lower limbs need to be calibrated with complicated steps before measurements, and the attachments of body also influence the normal motion and bring measurement errors caused by relative movements. This paper proposes a method using ground reaction forces (GRFs) and moments measured by a pair of wearable sensor shoes to estimate the joint angles of lower limbs for gait analysis. The wearable sensor shoes are combined with motion sensors to construct a wearable gait analysis system in our previous research. Based on the reliable measurements of the wearable system, neural networks are trained with GRFs or GRFs and moments as an input, and the joint angles as the output. The joint angles obtained from the trained models and the measurements of the motion capture system show good agreements. The correlation coefficients (R) are more than 0.7, 0.9, and 0.9 for the joint angle of ankle, knee, and hip, respectively, and the normalized root mean square errors are less than 8°, 9°, and 5° for the three joints, respectively. The comparison results show low performance of rectification of biases and distortion of waveforms. As a major advantage, the method to acquire kinematics using the wearable sensor shoes is relative simple and reliable compared with multi-inertial sensors on body segments.