Hybrid Distance-Based Framework for Classification of Embedded Firearm Recoil Data

Abstract

Internet of Battlefield Things (IoBT) system frameworks need to account for flexibility of design and unobtrusive implementation. Previous recoil based gunshot detection frameworks would require constant modification, or utilize bulky external sensors making their implementation into a true IoBT setting limited. By baselining the heterogeneous nature of firearm configurations, we examine differences in ammo and recoil characteristics utilizing the first application of Dynamic Time Warping (DTW) in an embedded recoil based gunshot detection framework. Statistical methods used for traditional Human Activity Recognition (HAR) frameworks would require constant refinement to account for differences in firearm systems across various platforms in an IoBT environment. Our proposed hybrid approach overcomes the limitations of both standalone DTW and statistical methods by combining features of both for the creation of an easily expandable gunshot detection framework. Our embedded sensor approach eliminates the obstruction caused by external sensor placement systems, providing a user friendly IoBT framework to expand upon in future research.