Gait Recognition using Histogram of Oriented Gradient and Self-Organizing Feature Map Classification in Variable Walking Speed

This paper proposes an vision-based gait identification system to recognize testers' identities in variable walking speed. This system uses silhouette-based images collected from a distance to recognize individuals' identifications without testers' cooperation, first and foremost, it adopts spatio-temporal gait representation method to build gait energy images (GEI) to obtain the characteristic of walking human because GEI method has the advantages of preserving temporal information and generating more abundant local shape features. Secondly, this system extracts histogram of oriented gradients (HOG) and low frequency wavelet based on discrete wavelet transformation (DWT) to decompose temporal space feature and reduce the dimension of the vector. Lastly, we utilize the self-organizing feature map (SOM) method to perform the feature vector classification and OU-ISIR (The Institute of Scientific and Industrial Research, Osaka University) gait database to verify the feasibility of the system. Experimental result shows this vision-based gait identification system can recognize testers' identity efficiently.