Date of Award


Document Type


Degree Name

Master of Science in Operations Research


Department of Mathematics and Statistics

First Advisor

Samuel A. Wright, PhD


This thesis addresses the final portion of a complete process for human gait recognition. The thesis takes as input information that has been generated from videotaping walking individuals and converting their gaits into numerical data that measures the locations of various points on the body through time. Beginning with this data, this thesis uses a variety of mathematical and statistical methods to create identifying signatures for each individual and identify them on the basis of that signature. The end goal is to achieve under controlled laboratory conditions human gait recognition, an identification method which does not require contact or cooperation with the individual and which can be done unobserved from a distance. Various mathematical models such as the construction of classifiers utilizing Minimum Euclidean Distance, Minimum Mahalanobis Distance and Quadratic Discriminant Functions are employed on both static and dynamic characteristics in order to fully analyze gait data for the purposes of identification.
This thesis starts with previously generated numerical data from a videotaped sequence of images of a subject walking across a room that contains the positions through time of a wide variety of different markers on the individual’s body. A MatLab program is initially written to convert the data into a usable format. A variety of mathematical techniques are then employed to generate several classifiers of an individual from a small set of gaits that can be used to identify their gait in any data set.

AFIT Designator