Gait Recognition Algorithm Research Based On Lower Leg Under45Degree Viewing Angle Of Video

Biometric recognition technology is a kind of personal recognition technology, andpeople focus on the security day and day, so we make this technology in-depth study. Gaitrecognition has notable advantages such as long-distance recognizability, difficult imitability,non-aggression and low requirement of clarity. It has a larger view field under45degreeviewing angle compared with0degree viewing angle. This paper presents two methods basedon gait recognition under45degree viewing angle.(1) A gait recognition method based on ankle joint motion trajectory and bending angleis proposed. First it obtains lower limb joint points according to each part of the body andheight proportion. It obtains the position coordinates of the toe by using skeleton algorithm.Then through the ankle joint motion trajectory obtain relative velocity. According to theposition relationship between joint points and toe extract, it can obtain bending angleinformation. The relative velocity and bending angle constitute the feature voter. Finallysupport vector machine (SVM) Classifier is used for the gait classification.(2)Gait recognition method based on velocity field and joint angle is proposed. Firstnormalization ankle joint motion trajectory and extracts its velocity fields from motiontrajectory of the ankle. Then vertical line with lower limb constitutes joint angles. The featurevector is made up of the velocity fields and joint angles and achieve gait classification. Finallyit is verified by using the ROC curve.The combination of ankle joint motion trajectory and joint angle becomes the featurevector. It reflects the time-space characteristic of gait motion and subtle changes of gaitpattern. It gives fully play to their complementarity of gait characteristics. Simulation resultson CASIA database show that the recognition rate obviously is higher than the recognitionmethod of only using single feature, it shows that this is an effective identification method.