Indoor positioning plays an increasingly important role in industrial spaces. Self-localizing autonomous machines are already well established and indoor positioning for pedestrians increases safety and productivity when interacting with (semi) autonomous machines. Positioning methods for pedestrians commonly use wearable or handheld sensors that rely on extensive calibration to tune gait detection heuristics. We present a method to automatically derive the target variables of these heuristics, stride length and stride orientation, for each individual user stride by using parallel location measurements during normal operation. We show that our method fits real world measurements of stride length. Additionally, we conduct simulations to verify the variance and bias of the derived stride length and orientation at varying accuracies of the reference position measurements. Our method provides a means for online annotation of raw data for pedestrian positioning, enabling both the online calibration of gait detection heuristics as well as data annotation for machine learning applications.