Automatic gait recognition using width vector mean

Gait recognition systems have recently attracted much interest from biometric researchers. In this work, we present an alternative gait representation of width vector profile. The proposed model-free gait representation, width vector mean, is defined by the arithmetic mean of width vector profiles obtained from a gait sequence. Different gait feature are extracted from the width vector mean such the downsampled width vector mean and the principal components of the width vector. To solve the classification problem, we use the Euclidean distance and a nearest neighbor (NN) approach. The Extensive experiments are carried out on the NLPR gait database to demonstrate the validity of the proposed gait representation.