Motion Recognition Under Simulation Prosthesis Vision

Eye diseases are one of the top three diseases affecting people's quality of life.China is the country with the most blind people in the world.The study of visual prosthesis provides blind people a viable visual alternative for the blind patients caused by Retinitis Pigmentosa and Age-related Macular Degeneration,.By electrically stimulating the optic nervous system,the prosthesis can generate phosphene in the visual center,which can provide an understandable visual perception for the prosthetic implant to restore some or all visual functions.The research on motion recognition under the visual simulation prosthesis has certain reference value for helping blind people to perceive things better in the external environment through the visual prosthesis and distinguish the human behavior.Research on motion recognition usually includes gesture recognition,motion recognition,gesture prediction and other directions.Gestures are not only an important body language,but vision-based gesture recognition technology is the main direction of human-computer interaction research.Therefore,in the research of artificial prosthesis vision,the recognition of gestures also has important research significance.The effect of gesture recognition depends to a certain extent on the quality of the image acquired by the camera,such as image resolution,whether it carries depth information,etc.,will affect the image quality.In this paper,the depth image acquired based on Kinect is studied for human motion recognition.Through the simple image processing of the acquired depth image,a motion recognition system is designed for the four movements of left shift,right shift,jumping up and down.The experimental results show that the motion recognition based on depth image can separate the human body from the background and achieve a good motion recognition effect.Next,through the different effects of color image and depth image on gesture recognition under simulated prosthetic vision,Kinect is used to obtain color image and depth image for gesture recognition research,and to explore a more appropriate image processing method under the simulated prosthesis vision.The depth image is captured by Kinect,and the obtained image is separated by threshold segmentation based on the depth distance using the Opencvlibrary function to separate the human body from the background,and then the median filtering is performed on the separated foreground image,and the basic morphological processing method is continuously applied to the image.The ground is subjected to corrosion and expansion treatment to remove image noise as much as possible,and to provide a better gesture posture image for subsequent gesture recognition.After the processed gesture image is pixlized,the gesture recognition experiment under the simulated prosthetic vision can be carried out.The experiment requires the subjects to perform a gesture recognition experiment with a total of ten digital gestures from 0 to 9 and a total of twenty-six letter gestures from A to Z at three resolutions(32×32,48×48,64×64).The results show that with the increase of resolution,the accuracy of gesture recognition is also increasing.At the same resolution,the gesture recognition rate under the depth image is generally higher than the gesture recognition rate under the color image,and at 32×32 resolution,the difference between the two is significant.