Gait Based Human Identification Research For Safety-Critical Environments

Persons' pass in and out should be mastered in government departments,army forces and other safety-critical places,and it requires to identify the moving human-being at a distance in most of the cases.At that time,other biological characteristics,such as face,iris,fingerprints,credentials and passwords are no longer available.However,gait recognition can be carried out effectively with low-resolution devices,which is a non-contact,non-offensive biometrics from a distance.Therefore,gait recognition is such an important technology,that can be used in safety video surveillance.Gait based human identification research for safety-critical environments pay attention to four major sections,namely,gait video sequences acquisition,gait detection,gait feature extraction and identification,and gait feature extraction is a key part in this paper.In gait video sequences acquisition section,we capture gait videos in the laboratory scenes,instead of existing databases.Capturing 20 individuals' gaits at a certain dip angle,the camera is fixed on a higher wall.The image preprocessing for gait videos is achieved,including gait video frame extraction,target detection,morphology and removal of redundant frames. Because of gait sequences captured in a dip angle,the segmentation of gait cycle images is ineffective.We use gait energy image(GEI) to denote gait characters.characteristics,which depending on,identification of individuals is achieved.The merit of GEI lies in that a human's movement sequence is characterized by a single frame image and the time information,profile information,frequency information and phase information are preserved.In the feature extraction section,two major categories of feature extraction methods are researched.One is gait energy image based on vector form reduction namely,principal component analysis(PCA),Kernel Principal Component Analysis(KPCA),linear discriminant analysis(LDA) and kernel linear discriminant analysis(KLDA).The other is gait energy image based on matrix form reduction namely,two-dimensional principal component analysis(2DPCA),the two direction two-dimensional principal component analysis((2D)~2PCA),the weighted two-dimensional principal component analysis(W2DPCA),the kernel two-dimensional principal component analysis(K2DPCA),two-dimensional linear discriminant analysis(2DLDA),two-way two-dimensional linear discriminant analysis(B2DLDA),2DPCA+2DLDA combination,K2DPCA +2DLDA with a aim to extract the eigenvectors with a biggest contributions to recognition eigenvector.We found a gait database including 20 individuals to evaluated to recognition performance and the speed of match through a single walking direction and a multiple walking directions experiments.Experimental results show that the method proposed in this paper,which making it possible that gait based human identification for safety-critical environments.Especially,gait feature extraction based on gait energy image matrix has e a good performance in the recognition accuracy and matching speed.It satisfied the demand of identification real-time and accuracy in intelligent safety video surveillance.