Utilizing 3D joints data extracted through depth camera to train classifiers for identifying suicide bomber

Abstract

Safety and security of humans is an important concern in every aspect. With the advancement in engineering, sciences, and technology (unfortunately) new methods to harm humans have also been introduced. At the same time, scientists are paying attention to the security aspects by developing new software and hardware gadgets. In comparison to the system level security, the safety/security of human beings is more important. Suicide bombing is one such nuisance that is still an open challenge for the world to detect before it is triggered. This work deals with the identification of a suicide bomber using a 3D depth camera and machine learning techniques. This work utilizes the skeletal data provided by the 3D depth camera to identify a bomber wearing a suicide jacket. The prediction is based on real-time 3D posture data of the body joints obtained through the depth camera. Using a comprehensive experimental design, a dataset is created consisting of 20 joints information obtained from 120 participants. The dataset records this for each of the participants with and without wearing a suicide jacket. Experiments are performed with the suicide jacket bearing 10- to 20-kg weight. Simulations are performed using 3D spatial features of the participants' body in four ways: full body joints (20 joints), upper-half of the body (above the spine base of the skeleton), 20 joints with 15 frames, and 20 joints with 20 frames. It is observed that 15 to 20 frames are sufficient to identify a suspected suicide bomber. The proposed framework utilize four classifiers to identify vulnerability of a subject to be a suicide bomber. Results show that the proposed framework is capable of identifying a suicide bomber with an average accuracy of 92.30%.