Adaptive Zero Velocity Detector for Pedestrian Navigation Based on Relative Test Statistics

Autonomous pedestrian localization involving complex motion modes is important for applications, such as firefighting and counter-terrorism in global navigation satellite system denied environments. Inertial navigation based on Zero Velocity Update is a crucial method, and adaptive zero velocity detector (AZVD) is a key part to improve performance. However, current AZVD still face challenges in adaptability to different motion speeds and complex motion modes (such as crawling), as well as high-computational load. To address these issues, this article proposes an AZVD that represents the generalized likelihood ratio test statistics within a time window as a ratio relative to the minimum statistic and sets thresholds based on these ratios (relative statistics). First, we theoretically analyze the method based on relative statistics. To ensure that the time window for data segmentation includes zero-velocity phases, and the minimum statistic corresponds to the zero-velocity phase, we divide the algorithm into two steps: 1) design a gait cycle (GC) detection suitable for single support motions and segment the data according to the GC. For motions where the GC cannot be detected, a designed time window is used for data segmentation and 2) for motions where steps are detectable, we restrict the zero-velocity phase to a specific range within the GC; for motions where the GC are undetectable, no restriction is imposed. Then, the threshold is set based on the relative statistics. Finally, through experiments involving multiple individuals and various motion speeds and motion modes, the adaptability of the proposed algorithm is validated.