Adaptive Threshold-Based ZUPT for Single IMU-Enabled Wearable Pedestrian Localization

Abstract

Without dependence on external anchors, the Micro-Electro-Mechanical System Inertia Measurement Unit (MIMU) allowing autonomous localization has promised great potential in wearable IoT applications, including kinematic analysis in sports, medical treatment, elderly care, and disaster rescue. However, the miniaturization of devices for unobtrusive sensing, algorithms minimizing the inherent accumulative errors of inertial devices, and the adaptivity of algorithms for various motion modalities are the key challenges. The removal of accumulative error in the continuous gait cycles with adaptive algorithms is a critical issue regarding localization accuracy, especially for low-cost devices. This investigation proposes an adaptive threshold-based Zero-velocity Update (ZUPT) algorithm to separate the timing of gait cycle phases and compensate for the residual velocity with a linear fitting approximation. The key contributions include: (1) a lightweight threshold-based zero-velocity detection algorithm to split the gait cycle phases of continuous walking; (2) a quaternion-based Extended Kalman Filter (EKF) algorithm to reduce the errors of the non-linear operations for attitude prediction; (3) a linear fitting method for compensating the residual velocity in each gait cycle of continuous walking; (4) the design of a miniature single-MIMU based foot-mountable wearable device and the corresponding experimental studies to verify the proposed methods. Results show that the relative error is less than 3.0% for 2D and 3D trajectories, and the tests with different locomotion patterns demonstrate the adaptivity of the proposed algorithm compared to its peers. The results show that the presented techniques are capable of handling accumulative errors for low-cost MIMU-based systems with good adaptivity.