Adaptive Zero Velocity Update Based on Velocity Classification for Pedestrian Tracking

摘要

Due to the drift of low-cost micro-electro-mechanical systems sensors and recursive computation, pedestrian tracking based on foot-mounted inertial sensors suffers from accumulative velocity and position errors. The well-known approach named zero velocity update (ZUPT) is able to minimize such errors by resetting the velocity to zero when the foot does not move with respect to ground. This time duration is called the still phase of a stride and can be detected by comparing the actual sensor data with a pre-defined threshold. However, when the velocity increases or decreases, the fixed threshold may be too small or too big for the correct detection of the still phase. Based on our previous study, this paper presents a novel adaptive ZUPT approach using an additional chest-attached accelerometer for consistent removal of accumulative errors even if the velocity changes. The information extracted from the chest acceleration is used to update corresponding threshold for the still-phase detection, and, hence, to determine error-free velocity and position information. The experimental results show that the accurate tracking trajectory can still be successfully obtained even if the velocity changes rapidly during the experiment.