Research On Application Of RGB-D Sensor In Upper Limb Rehabilitation Robot

Cerebral stroke is a common disease among the elderly, and it oftencauses hemiplegic paralysis. Nowadays, the increasingly ageing societymakes more and more countries attach importance to the prevention andtreatment of cerebral stroke. Over the past twenty years, upper limbrehabilitation robot has become a powerful tool for stroke rehabilitation,but now there remains room for its human machine interactive function tobe improved. In recent years, somatosensory interaction devices based oncomputer vision technology is popular among more and more people, andone of such device, known as "as great as the mouse", is Kinect, which cancollect the RGB-D data of the scene, and implement digital exoskeletontracking of a human body by a professional level chip. This paper mainlystudies the application of the RGB-D device Kinect to the upper limbsrehabilitation robot.According to different phases in the rehabilitation process of patientswith hemiplegia, this paper is divided into two application scenarios: one iswhen the patients are in a stage of weakness and spasm, in order toimprove the patient’s capacity of independent home rehabilitation training,healthy limb’s trajectory is obtained from the physician or the patienthimself through Kinect, then after Kalman filtering, a mirror of trajectoryis generated, and its joint Angles are directly transmitted to drive theexoskeleton robot to exercise the paretic limb; This paper designed a4-DoF exoskeleton robot, which was combined with Kinect and asemi-physical simulation experiment has been carried out. The otherscenario is when patients possess more voluntary movement function inthe convalescence. In order to improve the interest and compliance of thepatients during trainings, this paper designed a gesture based HMI system,achieving a “Machine to the Human” effect; Firstly, this paper studied theapplication of Kinect to the image-based visual servo (IBVS) system, which consequently realized a Hand-Shake style automatic docking effect;Secondly an innovative method to obtain the position indicated wasproposed, and this method is promising to be applied to the detection oftarget and the development of rehabilitation game.The semi-physical simulation experiments for each applicationscenario show that the application of RGB-D sensor is effective inimproving the upper limb rehabilitation robot’s human-computerinteraction function, and in enhancing hemiplegic patients’ ability ofself-initiating the training at home and also their compliance whentraining.