Study On Gait Recognition Method Based On Fusion Of Lower Limb Angle And Plantar Pressure Distribution Features Based On Vision And Tactus

With the rapid development of computer vision,pattern recognition and artificial intelligence technology and urgent need for public security in modern society,researchers pay attention to the safety of public places,such as banks,airports,customs.The traditional identification method is gradually replaced by biometric identification technology due to its advantages of uniqueness,universality and difficult to be lost and forged.Gait recognition is a new biometric identification technology,which attracts wide attention of researchers due to its advantages of inconspicuous recognition at a relatively far distance,difficult to be hided and disguised.It has good application prospect in the fields of intelligent monitoring and medical diagnosis etc.Gait recognition is divided into gait recognition based on vision and gait recognition based on tactus.Gait recognition based on vision is susceptible to interference factors such as environment,perspective.Compared with gait recognition on vision,Gait recognition based on tactus is insusceptible to external factors.But the study on gait recognition based on tactus is in the initial stage and the description of gait tactile feature is unideal,leading to low recognition rate.Data fusion can greatly improve the recognition rate through features correlation.Therefore,this paper presents a gait recognition algorithm based on fusion of vision and tactus.Gait recognition process mainly includes three aspects of preprocessing,feature extraction and classification.Preprocessing generally includes moving objects detection and periodic analysis.Using background subtraction method segments moving objects from background image due to the characteristics of database with static camera,simple and single environment.Using the aspect ratio of toros makes periodic analysis of gait motion.In order to reduce the dimension of features and improve operation speed,the key frames of gait is extracted.Aiming at the problem of self-occlusion and shadow,a method of automatically positioning the joint points of the human body is proposed.The method can automatically position the joints of lower limbs,and the algorithm is simple and fast,which is only affected by the outer contour points of surrounding joints.Self occlusion and shadow has little effect on the results.According to the knowledge of human anatomy and the characteristics of the body structure,the angles of the lower limb are extracted.According to the characteristics of the plantar structure,the foot is divided into four parts.The features of maximum pressure points and coordinates,the center of pressure,the aspect ratio of torso are extracted.In view of the problem that the gait recognition rate based on single feature is not high,a algorithm baded on fusion of lower limb angles features and characteristics of plantar pressure distribution in the feature layer is proposed.The information of lower limb joint angles contains more gait movement characteristics and gait patterns of subtle changes,which fully reflects the gait of walking.The information of plantar pressure distribution directly reflects the change of plantar pressure during walking.The features of lower limb joint angles and plantar pressure distribution are fused in the feature layer to form a new feature vector.The nearest neighbor method and support vector machine are used to classify gait.The experimental results show that the gait recognition rate based on fusion of lower limb joint angles and the plantar pressure distribution features is higher than the gait recognition rate based on the single feature.The feasibility of this method is verified.