A Frequency Domain Analysis Based Gait Feature Representation

A method for representing gait feature based on frequency domain analysis is proposed. For a gait represented by a sequence of binarized silhouettes, its frequency domain feature is extracted following period detection, contour extraction and unwrapping, distance normalization, 2D Fourier transform, feature frequencies computation and some other procedures. Nearest neighbor classifier is used to test its usefulness for gait recognition purpose, and experiment result, compared with previously conducted research, shows that the proposed gait feature representation is soundly effective.