An adaptive gait event detection method based on stance point for walking assistive devices

This paper presents an adaptive and fuzzy logic-based gait event detection method for wearable assistive devices. A conventional and straightforward way to detect gait events is to utilize gyroscope measurements in the sagittal plane for time-series pattern recognition (positive peaks and negative peaks) based on a predefined threshold. This approach works well in the biomechanics analysis while it may have difficulties adapting to the changes in human walking speed for wearable robot applications. To tackle the above issue, first, we keep updating the detection threshold according to the last stride information. Second, we detect the stance point (zero-velocity point) as an indicator to distinguish between the heel strike and toe off events by combining the information about the foot angular velocity and acceleration. A method to construct a fuzzy membership function is also proposed via a series of moving intervals from foot acceleration data. Validation of the proposed gait event detection method using force plates showed that the method obtained high detection accuracy (F1-score = 0.99) for healthy subjects with and without the robotic support limb (RSL).