Study On Gait Recognition Method Based On Visual-tactile Features Fusion

Biometric authentication uses inherent characteristics to recognize the identity of the person. Because of its uniqueness and stability, and so on, biometric authentication has become one of the hot issues of concern. Compared to other biometric identifications, gait recognition has the advantage of long distance and non intrusive, and aroused more and more attention in the field of recognition.Gait recognition is mainly divided into visual recognition and tactile recognition. Visual recognition is better than tactile recognition, but it is limited to light and clothing. On the contrary, tactile recognition is not influenced by light and shade. Therefore, the study of gait recognition with visual-tactile fusion is of great significance. Based on research on gait recognition, a gait recognition based on visual-tactile multi-feature fusion has been proposed. The specific achievements are as follows:This paper described the background and significance of the subject and the status of the domestic and international development were summarized. Based on gait kinematic analysis, lower extremity mechanics model has been established. Joint angle, angular velocity, angular acceleration of the lower limbs has been calculated. The feasibility of identification was analyzed in theory, specific identifying characteristics were determined.Gait recognition based on visual process includes detection of moving target, preprocessing, feature extraction, classification and recognition. Gait motion pictures were detected by using visual image processing method. Gait image sequences were obtained by morphological processing and key frame extraction. Varying pattern of leg angle and the changing rule were obtained by using measurement method. Then the representative identification features were selected as the characteristic of biometric authentication. According to the location of the hip joints and knee-joints, the rotation angle, angular velocity, angular acceleration of hip and knee were obtained. Plantar pressure curves were measured by using plantar pressure force plate. Then feature points on the curves were analyzed, the pressure and time parameters on the curve were obtained, these parameters were used as visual-tactile features.On the basis of visual-tactile and the fusion theory research, the fusion information was identified in the feature level. Under the inspiration of the biology, neural network classifier of visual-tactile information was constructed. The characteristics of artificial neural networks were analyzed and the various functions were determined. The final network was trained by inputting samples. The simulation results showed that the right rate of visual-tactile fusion was higher than the visual or tactile recognition. The feasibility of fusion method was proved.