Our team has a decade long history of research on countersniper systems. Our most significant contributions to the field include:
The disadvantage of our first approach is its static nature and the need for many sensors. Hence, we developed a soldier-wearable version that required only a handful of sensors. To compensate for the small number of sensors, each sensor was equipped with four microphones forming a tiny microphone array. As such, a single sensor alone was able to locate the shooter. Multiple sensor nodes still cooperated to provide much better accuracy and robustness as well as trajectory estimation and weapon classification.
The following picture shows a 100 m shot localized by the network (trajectory: black, shooter location: black dot). The weapon and caliber is also shown on the left. The white arrows indicate that 6 out of the 10 sensors were able to localize the shooter alone, that is, they relied only on their own detections using their microphones with 10 cm separation.
More details can be found in our Mobisys 2007 paper.
The motivation behind the mobile phone-based approach is simply the fact that most people including soldiers already carry smartphones. So why not utilize it? While the entire approach could be implemented on phones since they already have a microphone, a GPS, WIFI or 3G networking, a display and enough computing power to carry out the sensor fusion, it would not be too accurate due to various reasons and continuous sampling of the microphone would drain the battery fast. Instead we developed a costum bluetooth headset that doublses as a single-channel sensor node. It only needs to utlize that phone, when there was a shot detected.
The challenge of course is the fact that we now only have a handful of sensors and they only have a sinlge microphone each. Therefore, we developed a new sensor fusion approach that utilizes the shockwave length for miss distance and trajectory estimation. The time of arrival of the muzzle blast and the shockwave, in turn, make it possible to localize the shooter and classify the weapon used. This works surprisingly well as described in our Sensys 2011 paper.
We gratefully acknowledge the generous sponsoship of DARPA, ONR, ARL and Databouy Inc. that made our research possible.
For more information contact Akos Ledeczi