Fusion Recognition of Micro-gait and Micro-expression Based on Microscope Technology

  • Beiji Zou, Lingzi Jiang
Keywords: Gait Recognition; Face Recognition; Multiple Biometrics; Microscopic Techniques

Abstract

In order to maintain the advantages of gait recognition, overcome the shortcomings of single biological feature recognition and improve the recognition rate of long-distance identity, a method of gait and side face fusion recognition at feature level is proposed. Gait recognition has become the most potential emerging hot spot in this field because of its advantages in remote range identification. But at present, the research in this field is still in the primary stage of laboratory. It is still an arduous task to develop an accurate, reliable, robust and practical gait recognition system. This paper mainly studies the recognition methods of two biometrics based on face and gait in walking video. This paper describes a three-dimensional reconstruction technique based on two-dimensional fused image under microscope, which can restore the three-dimensional features of two-dimensional fused image. The three-dimensional simulation system is a real-time system which converts the real scene into computer virtual reality. Taking the traditional microscopic biological experiment as an example, a three-dimensional simulation system of microscopic biological experiment is constructed. An adaptive weighted fusion method based on decision level is proposed for face and gait under multiple angle videos. Firstly, the feature extraction and dimensionality reduction of gait energy map and side face map are carried out respectively by using the principal component analysis of binary image matrix. The initial eigenvector matrix is obtained, and the initial eigenvector matrix is vectorized and combined to obtain the combined eigenvector. Then multi-discriminant analysis is used to fuse the combined feature vectors to obtain the fused feature vectors of gait and face. Finally, the nearest neighbor method is used for personal identification. The above method is verified by CASIA Dataset B gait database. The results show that the method improves the accuracy of identification, verifies the effectiveness of the method, and provides a new method for multi-biometric recognition.

Published
2019-04-13