Analysis of Gait Biometric in the Frequency Domain

Person reidentification is an essential task in video surveillance. Tracking a person in different cameras, backgrounds and views are important for human identification. Human gait is used to identify and recognize a person from a distance. Features are extracted from the series of images using background subtraction techniques. The extracted features are dimensionality reduced, from higher dimensionality to lower dimension, and classified using different classifiers. In this work, the Gait Energy Image, Gait Energy Image with Discrete Fourier Transform, and Gait Energy Image with Gabor filter features are extracted from people's normal walking, carrying a bag, carrying a suitcase, and wearing a coat. The experiment shows frequency domain methods has an encouraging recognition performance compared to the traditional Gait Energy Image method.