IoT audio sensors flag underreported gunfire
Police in Dayton, Ohio, installed audio sensors in neighborhoods that were generating a large number of complaints about gunshots, expecting the internet-of-things technology that helps locate the origins of gunfire would speed response, improve evidence collection and better document the number of shots fired.
The ShotSpotter audio sensors, however, detected far more gunshots than were reported by neighborhood residents. Shortly after the system was installed in mid-December, the technology captured over 40 incidents of gunfire in one neighborhood, but only one or two were ever reported to authorities, according to a report in the Journal News.
The strategically placed ShotSpotter audio sensors use audio triangulation to pinpoint -- within 100 feet -- and timestamp gunfire location. Machine-learning algorithms compare the acoustic waves with an extensive gunfire database to confirm whether the sound is actually gunfire and if multiple shooters or automatic weapons, for example, are involved. After trained analysts quickly validate the report, the system then alerts law enforcement and emergency responders -- in less than 60 seconds, company officials said.
“It is estimated that 80% to 90% of the time a gun is fired, there’s no call to 911,” said ShotSpotter CEO Ralph Clark told the Journal News. “As a result, there is no response.”
Dayton's system flagged dozens of incidents in the last two weeks of December, the Journal News reported. In one case, no one reported about a dozen shots, in another a shooting victim was discovered, an in a third, police found an armed suspect after responding to an alert.
ShotSpotter is used by nearly 100 cities, including San Diego and Louisville, Ky. Those two cities are also considering mounting the sensors on drones to increase the precision of the triangulation and provide a video feed of the area where shots were fired, without the cost of building out a dense network of cameras and sensors.
Connect with the GCN staff on Twitter @GCNtech.