Uncertain speed pedestrian gait classification

The gait classification is the first step of behavior understanding. In the process of classification, an important cue of pedestrian is silhouette. The paper proposes a complex number notation to indicate a pedestrian's contour. A set of standard gait image contours and their reverse silhouettes are stored in base according to the different gait. A distance method is used to get the similarity of two shapes. In the classification stage the minimal standard deviation algorithm is proposed. The incoming video contours are compared with every motion sequence, and then find out the minimal value of every motion. Finally the pedestrian classification is deduced by the minimal standard deviation. In this paper two groups of experiments are done with a high recognition rate. 1548-7741/Copyright © 2008 Binary Information Press October 2008.