Gait-based human identification using appearance matching

In this chapter, we present an appearance-based approach for recognizing human gait. Given the gait video of an individual, the images are binarized and the width of the outer contour of the silhouette of that individual is obtained for each image frame. Several gait features are derived from this basic width vector. Temporally ordered sequences of the feature vectors are then used to represent the gait of a person. While matching the feature templates for recognition, dynamic time-warping (DTW), which is a nonlinear time-normalization technique, is used to deal with naturally occurring changes in the walking speeds of individuals. The performance of the proposed method is tested on indoor as well as outdoor gait databases, and the efficacy of different gait features and their noise resilience is studied. The experiments also demonstrate the effect of change in the viewing angle and frame rate of data capture on the accuracy of gait recognition.