Identity Recognition System based on Walking Posture

It is well known that identity recognition system is one of the key aeras of computer vision research. Meanwhile, the mainstream identity recognition systems often use the features of people’s face or appearance to realize identification, but the features of face and appearance are very easy to be changed. On the contrary, the features of walking posture are steady for specific person, because walking posture have a strong connection with people’s habit and their biological characteristics. Therefore, in this paper a deep learning algorithm based on walking posture is proposed. The proposed model use OpenPose algorithm to get a series skeleton feature pictures of people’s walking posture through video data, then use CNN to extract the information of pictures and use RNN to analyze this information to get identifications.