Research On Prediction Method Of Rehabilitation Robot Assisted Patient Standing-up Trajectory

Many elderly people are unable to stand up independently due to physical degradation and other reasons.And,the existing artificial nursing has many problems such as nursing shortage,without nursing at any moment and so on.The emergence of rehabilitation robot provides the possibility to solve these problems.The existing assistive standing-up robots are able to complete the assistive standing-up task under certain conditions.However,there is less research on what kind of trajectories should be used in the process of standing up.There are two main types of assistive standing-up trajectories used by existing assistive standing-up robots.One is fixed trajectory,another is adjustable trajectory.These two kinds of trajectories cannot meet the requirements of users in different physical conditions or different postures.In this thesis,we study the algorithm of auxiliary joint trajectory prediction based on machine vision and support vector regression.The trajectory predicted by this algorithm meets the requirements of different standing postures in the different sitting postures.By analyzing the process of human standing up,we found that the human standing up movement is completed by the shoulder,hip,knee and ankle and at the same time,the joint distance also has an effect on the rise process.Therefore,the mathematical model of the human body standing up motion is established by the joint parameters and time serial numbers.In order to obtain the accurate mathematical model of joint movement during the process of human body standing up in different heights and sitting postures,we need to train the established model.The data of human standing up joint movement were obtained by machine vision and image processing.Kinect depth camera collects human body standing up motion images.The collected images are preprocessed to reduce the influence of noise and edge on feature extraction.SIFT feature extraction algorithm is used to extract the features of the marker points,and the RANSAC algorithm is used to remove the false matching points.At last,the coordinates of the marker points are obtained by a homography matrix.The basic principles of three common machine learning regression algorithms are analyzed and the learning ability and predictive ability of the three regression models were verified by a series of comparative experiment.It is found that the support vector regression algorithm has better performances.The support vector regression algorithm and the parameters of human joint motion are used to train the mathematical model of the standing up motion.A trained model is used to predict the motion trajectories of a fresh sample and compare with independently standing up trajectories.Finally,the auxiliary standing robot completes the standing task according to the predicted trajectory.The experimental results show that the proposed algorithm is feasible.