Firearm Detection Using Wrist Worn Tri-Axis Accelerometer Signals

Abstract

Gunshot detection traditionally has been a task performed with acoustic signal processing. While this type of detection can give cities, civil services and training institutes a method to identify specific locations of gunshots, the nature of acoustic detection may not provide the fine-grained detection accuracy and sufficient metrics for performance assessment. If however you examine a different signature of a gunshot, the recoil, detection of the same event with accelerometers can provide you with persona and firearm model level detection abilities. The functionality of accelerometer sensors in wrist worn devices have increased significantly in recent time. From fitness trackers to smart watches, accelerometers have been put to use in various activity recognition and detection applications. In this paper, we design an approach that is able to account for the variations in firearm generated recoil, as recorded by a wrist worn accelerometer, and helps categorize the impulse forces. Our experiments show that not only can wrist worn accelerometers detect the differences in handgun rifle and shotgun gunshots, but the individual models of firearms can be distinguished from each other. The application of this framework could be extended in the future to include real time detection embedded in smart devices to assist in firearms training and also help in crime detection and prosecution.