Front View Gait Recogniton Research By The Fusion Of Moving Gestures And Gait Energy Image

The wide application of network promotes the level of social information with the rapid development of science and technology,people have raised their awareness of self-protection.It has been an important issue for society to observe that how to avoid unnecessary loss of property and information,the study of biological recognition technology has come up with a solution to this problem.Gait recognition is the only biometric technology that can recognize people within a certain distance compared with other identification technologies,it has been a research hotspot of personal identification.The video captures many pedestrians' gaits from positive perspective in actual monitor but correlative research reports are few,so the gait in front-view also has important research significance.This paper focuses on gait in front view.In this paper,we study the front-view gait videoed by camera,including indoor and outdoor environment.This paper divides the process of gait recognition into four steps: the target detection,cycle extraction,features extraction and classification recognition,which will be studied respectively.For target detection,three kinds of common methods of target extraction are experimented and analyzed preliminarily.The most effective method will be adopted after comparison,after image binarization,morphology denoising and a series of post-processing method,the target areas are more integrated and the body contour become clearer and more accurate.This paper puts forward three kinds of cycle detection methods by analyzing the cycle characteristics of different areas for cycle detection: Based on the feature of the holistic swinging body;Based on the feature of changing area of lower arms;Based on the feature of changing area of lower limbs.It is the most accurate and fastest method that based on the feature of the holistic swinging body after experiments and comparison,which will be adopted to detect the cycle of gait.Feature extraction is the key step of identification process,the selection of features is very important.In order to figure that recognition result will be not ideal and lack of robustness if only single feature adopted for recognition.This paper describes gait silhouettes in a period by gait energy image(GEI)and extracts features based on gait energy image.At the same time,this paper puts forward to use cycle characteristics of moving gesture as one of gait features.This paper proposes a fusion method of three kinds of characteristics in pairs for recognition and uses support vector machine(SVM)as classifier.The specific experiments based on gait database provided by Chinese Academy of Sciences have validated the feasibility and validity of the proposed recognition algorithm.