Vision-based head posture control wheelchair system research

In terms of interactive equipment with traditional man-machines, now it can directly realize human-computer interaction in the field of computer vision. In order to solve the problems of daily operations of disabled people, build a wheelchair control system combined by soft hardware. Users only need to use simple devices to complete the goals of daily operations. The control system mainly uses the real-time facial key point detection method of MediaPipe. Use a USB camera to achieve precise head posture estimation. For the error ratio of the algorithm, the average error ratio of the partial yaw angle obtained in the end is (0.0579), the average error ratio of the pitch angle is (0.0518), and the average error ratio of the roll angle is (0.0294). The maximum error is the yaw angle measurement angle (0.065), The overall error is within 1 degree to meet the requirements of experimental accuracy. The real-time motion control task of the wheelchair is completed through the non-contact human-computer interaction mode of camera head attitude estimation, and the wheelchair drive control is realized in the ROS system. Through experimental verification, the user successfully reaches the destination in a complex environment through head attitude control in a visual way, providing a solution to the human-computer interaction problem of the wheelchair; This can assist those with special disabilities, making it more user-friendly for real-time control of wheelchairs.