Smartphone-based mobile gunshot detection

David Welsh, Nirmalya Roy

Abstract: The ability to detect gunshots can provide someone with invaluable information in various circumstances. For the military and public servants, detecting gunshots can help save lives and potentially target offenders. People participating in shooting sports as beginners or professionals can also benefit from the use of sensors for improving their reaction and self control during training. Most current methods for gunshot detection require expensive devices that are purpose built or developed and often only examine one or two features of the gunshots such as the sound, recoil, or visible flash. Using the current sensors built into smartphones, 15 samples utilizing 10 different sensors are used examine how gunshot detection can be performed through the use of simple sensors. This extension of human activity recognition resulted in gunshot classification accuracy ranging from 0.0% - 99.7% with an average of 86.6%. Understanding how simple sensors respond to gunshots can provide simple and easy accessibility to new detection methods and ample opportunities to improve this potential field for various personal and smart city applications such as crime detection, policing, gunshot violence monitoring and control in the community.