The Key Technology Research, Video-based Motion Analysis

This paper presents a action recognition-based methodology for action 2recognition of rehabilitation therapy of human.we discuss the key technology of action recognition which is record by video. Rehabilitation therapy is an important part of rehabilitation medicine and is the important method to restore patient health and function. Rehabilitation is also a component of comprehensive treatment.The main means of rehabilitation is exercise training, Currently rehabilitation physicians,rehabilitation therapists and clinical staff composed of the relevant bodies to implement the rehabilitation treatment,sometimes with auxiliary sensors.These are the traditional non-smart way,into the years of the development of technology has emerged a number of different types of intelligent rehabilitation system for patients with different types,for example, intelligent physical rehabilitant device and intelligent multi-functional electrical rehabilitant bed.In short the whole world system of rehabilitation has made tremendous progress. Human motion analysis is the frontier field of computer vision,the core is use of computer vision technology to identify people from image sequence,and to understand and describe their behavior. Motion analysis is mainly related to pattern recognition, image processing,computer vision, artifical intelligence and other academic knowledge.It is have broad application prospects in virtual reality,visual surveillance,and Perceptual Interface.We main concern about identification problem of Rehabilitation therapy, the research content is the foreground extraction,motion tracking and motion understanding. Mainly done as follows:1.analysis of several different algorithms in the foreground extraction results.2.Select the characteristics of human star-skeleton approach to tracking human motion and use HMM to recognize the motion.Proposed using particle swarm optimization vector quantization star skeleton features,and through experiments proved the method is superior the traditional LBG algorithm to generatethe feature vector codebook.Finally,this paper summarizes and analyzes the related work.