Minimizing Validation Artifacts in Azure Kinect Caused by Marker-Based Systems Using Synchronized Markers

The Azure Kinect is a human pose tracking device suitable for a wide range of applications. To evaluate its accuracy, it is frequently compared to marker-based motion tracking systems, which are regarded as the gold standard in human pose tracking. However, the infrared light emitted by this system, as well as the reflective markers attached to the body, significantly impair the Kinect’s ability to track the body, thus hindering a meaningful validation of the system. To address this issue, we developed a novel set of active markers synchronized with the Kinect’s shutter, effectively eliminating these sources of interference. The system’s efficacy was assessed through a study involving ten participants, who performed ten distinct exercises using both the newly developed synchronized markers and the standard Vicon markers. When using the synchronized markers, as opposed to the standard ones, the median positional error across all joints decreased from 2.4 to 0.96 cm, and the Pearson correlation across all joints increased from a mean of 0.48 to 0.70. The results demonstrate that the synchronized markers do not cause interference and are not detrimental to the Kinect’s capacity for accurate and reliable individual tracking. Therefore, the synchronized markers are considerably better suited for validating the Kinect against a marker-based reference system. To enable others to benefit from these improvements, we have published an assembly manual alongside the necessary resources.