We present a real-time human body-segment (e.g., upper limbs) orientation estimation scheme in rider-bicycle interactions. The estimation scheme is built on the fusion of measurements of an un-calibrated monocular camera on the bicycle and a set of small wearable gyroscopes attached to rider’s upper limbs. The known optical features are conveniently collocated with the gyroscopes. The design of an extended Kalman filter (EKF) to fuse the vision/inertial measurements compensates for the drifting errors by directly integrating gyroscope measurements. The characteristic and constraints from human anatomy and the rider-bicycle interactions are used to enhance the EKF performance. We demonstrate the effectiveness of the estimation design through bicycle riding experiments. The attractive properties of the proposed pose estimation in human-machine interactions include low-cost, high-accuracy, and wearable configurations for outdoor personal activities. Although we only present the application for rider-bicycle interactions, the proposed estimation scheme is readily extended and used for other types of human-machine interactions.

This content is only available via PDF.
You do not currently have access to this content.