A robust step length estimation system for human gait using motion sensors

Abstract

Wireless motion sensors offer a non-invasive, low cost solution for daily human activity monitoring, which is critically important for the diagnosis and rehabilitation of neurological diseases. However, an accurate and robust estimation method for the step length, one of the most important lower body metrics, is still lacking. In order to tackle this problem, we developed a new robust step length estimation system called the Pose Invariant (PI) method using ankle mounted motion sensors. Using the inverted pendulum model, the traveling distance within each step can be calculated by multiplying the leg length and the sine of the leg's orientation change within each step. Walking data from 9 adult subjects was collected and processed to validate this method. On average, a 3.69% absolute error rate was achieved. In addition, the robustness of this method compared to the non-ZUPT method was shown by an additional experiment over 3 adult subjects.