Design Of Cognitive Rehabilitation Robotic System Based On Machine Vision

With the intensive aging of China's society,there is a significant increase in the number of people with needs for rehabilitation.In this thesis,artificial intelligence technology is adopted,such as machine vision,to assist in the rehabilitation of mild cognitive impairment(MCI).MCI is a kind of cognitive disorder syndrome,and it has been proved that,to a certain extent,MCI can be slowed down or even cured by human intervention.However,due to the lack and uneven distribution of related medical resources,the recovery processes of patients are often not in time or in place,and lacks continuity and effectiveness.In this paper,a machine vision based robotic system for assistance of cognitive rehabilitation is designed to improve the recovery efficiency of the disease such as MCI.First of all,based on Wechsler Adult Intelligence Scale(WAIS),the cognitive abilities of MCI are obtained.The block design rehabilitation strategy is selected as the main rehabilitation method,assistance is carried out according to its process,and the "one-to-many" assistant rehabilitation mode has been proposed.Secondly,target detection,image matching and parameter extraction are accomplished in the vision part of the system.The Fourier descriptor is adopt to convert image information from time domain to frequency domain,and target detection is accomplished by comparing the descriptors of reference objects and objects to be tested.And then,the Module-Shift Matching algorithm is proposed.Using the complex field information of images,modulus part of complex series of target image is shifted so as to realize the rotational transformation,the difference of the modulus length at the same phase angle between the two objects are compared to determine the angle change without considering the starting point of the edge of image.Also,support vector machine(SVM)algorithm is adopted,the features of the spatial objects are learned,and three-dimensional target detection is carried out according to the classifier obtained from learning.Then,two kinds of software interfaces were designed for physicians and patients,respectively.The physician UI is developed for the overall control of the system,and the patient UI is mainly used for giving proper assistance during the rehabilitation process.Based on this,the "one-to-many" assistant rehabilitation strategy is designed: block design rehabilitation tasks were assigned by one physician to multiple patients through their software,which would be sent to software windows of different patients,each patient is equipped with the same rehabilitation system devices.If the patient encounter difficulties,firstly,based on machine vision,highlight hint will be carried out through the interface.If the patient is still unable to complete the task,the robotic arm will be started to show a demonstration of block design task for patient.Next,a complete assistant rehabilitation system is built adopting six degrees of freedom(6-DOF)robotic arm,a monocular camera,a block rehabilitation platform and a display device.The movement and obstacle avoidance strategy of the robotic arm is analyzed.Finally,a complete test with simulated patient was performed in the model machine of rehabilitation system.The results of the model machine test fully demonstrate the feasibility of the assistant rehabilitation strategy and the high efficiency of the assistant rehabilitation robotic system.To a certain extent,patients can achieve self-help rehabilitation.