Research on Key Technology of Multi-View Gait Recognition Based on Neural Network

  • Chen Ma , Lei Lei

Abstract

Gait recognition can be performed by taking videos at a distance, without using too many local details, overcoming some of the limitations of current biometrics. Gait recognition technology has obvious advantages, can overcome the shortcomings of current face recognition and other methods, and can be widely used in complex scenes. Improving the accuracy of gait identification will help to improve efficiency and reduce workload, especially in the efficiency of pedestrian identification and authentication at stations. The difficulty to be solved at present is the cross-perspective problem, also known as the multi-perspective problem. This paper will focus on this problem, through deep learning and other methods to carry out research, the main work is as follows: (1) The current research status of gait identification and multi-perspective issues at home and abroad was investigated. (2) Research on perspective conversion model based on generative adversarial network. In this paper, by training a generative adversarial network, the network can convert gait sequences from other perspectives to a unified perspective for recognition. (3) Research on gait recognition model based on human posture. This paper uses the human body joint length, joint angle, angular acceleration and other features extracted from the human body pose coordinates to form a feature matrix for identifying pedestrian identity.

How to Cite
Chen Ma , Lei Lei. (1). Research on Key Technology of Multi-View Gait Recognition Based on Neural Network. Forest Chemicals Review, 1683-1702. Retrieved from http://forestchemicalsreview.com/index.php/JFCR/article/view/1036
Section
Articles