Document Type



Active foot contact (absence of a braking impulse) during the acceleration phase of athletic sprinting is associated with the motion of the foot before touchdown (TD). Since the identification of braking impulses through force plate measurements is cost-expensive, the aim of this study was to develop a machine learning algorithm to predict active foot contact occurrences based on ankle-mounted accelerometer measurements. Ten recreationally active athletes (three females, seven males) performed 30 sprint-block-starts each, which were used as input to the machine learning model. Model performance was assessed by the AUC for both validation (AUC = 0.96) and testing (AUC = 0.94). It is therefore possible to predict active foot contact occurrence by a machine learning algorithm solely based on ankle-mounted accelerometer measurement data.