In the midst of an outdoor celebration in Baghdad, a young woman who had just cleared a checkpoint joined the celebration and blew herself up, killing 25 people and injuring 28. Individuals who detonate bombs concealed on their bodies are an all-too-familiar scenario in the Middle East. There is also a real potential for this scenario in the West.
While vehicle bombs and IEDs are the primary method currently used to spread terror in Iraq and attack U.S. convoys, U.S. warfighters are also starting to face suicide bombers. To counter this growing concern, an inexpensive suicide bomber screening system that could screen individuals approaching a checkpoint while still 500 to 1,000 feet away is needed. The Center for MASINT Studies and Research (CMSR), in collaboration with others, is investigating video-derived gait characteristics associated with individuals carrying a bomb on their body. The biometric question is whether human gait signatures of individuals who are carrying no visible load on their bodies are distinct from those signatures of individuals with no load. The current challenge is to extend the National Institute of Standards and Technology (NIST) Human ID at a Distance gait program started in 2001. That program concentrated on recognition capability of outdoor data with the study of covariate factors such as lighting, footwear, apparel, surface conditions, and shadowing.
An outdoor gait track with a cross-over platform is being used, along with weighted vests and volunteer human subjects, to create a rich database for signature development. Two feature approaches are being pursued to address the problem. The first is a model-free analysis of silhouettes being led by Major David M. Kaziska, Assistant Professor of Mathematics. The initial results from Dr. Kaziska's shape-based approach were encouraging. For each subject, training and test gait templates were computed. Persons in the test set are identified by finding the nearest match in the training set, using a distance metric between the shapes in the gait templates at comparable time points. This method was applied to preliminary data in the integration of a sensor package for identification of radical extremists (INSPIRE) project by attempting to determine, for an individual, whether or not that individual was wearing a pack on the torso. Video was shot of subjects with and without the pack, and training and test templates were computed. In this initial study, six of eight matches were correct.
The second approach is a model-based analysis being led by Dr. Kimberly Kendricks. Dr. Kendricks will assist with efforts to develop a set of characteristics in gaits that are noticeable at a distance of 75 to 100 feet to determine whether or not an individual may pose a threat. Thus far, Dr. Kendricks has developed a mathematical model describing arm-swing movement and verified it by using Groebner basis theory to discover a Groebner basis that yields a one-to-one correspondence of solutions found through the forward and the inverse kinematic problems.
With continued effort, the CMSR hopes to develop a viable solution to this grave problem.
