Lift movement detection with a QDA classifier for an active hip exoskeleton

Abstract

To provide assistance with an active exoskeleton, the control system of the device has to automatically detect the onset of the user’s movement and provide timely assistance, according to the recognized movement. In this paper, we present an algorithm designed to detect the lift movement with an active pelvis exoskeleton, based on a quadratic-discriminant-analysis classifier combined with a rule-based algorithm. The algorithm relies on sensory information acquired from the sensory apparatus of the exoskeleton, without needing additional sensors to be placed on the user’s body. The algorithm was validated in experiments with seven healthy subjects. Participants were requested to execute different actions, i.e. lift and lower a load, stand up, sit down and walk, while wearing the exoskeleton. On average, the algorithm showed an accuracy of 98.7 ± 0.6% in recognizing the lift task; such performance make it suitable for use in real application scenarios.