Apparently with 100 miles of listening coverage ShotSpotter claims a June average of 85 gunshots detected per night from 30 cities including Chicago, Minneapolis, Washington and Oakland.

Their technology is based upon earthquake monitoring systems (and we know how accurate those are). It uses a network of listening sensors to identify sound wave patterns, triangulate them, and then notify nearby law enforcement. The company marketing page makes some bold claims:

…ShotSpotter systems are not fooled by noises which sound like gunfire but are misleading (like car backfires, firecrackers, etc.). Similarly, the technology filters out echoes and other acoustical anomalies. Using a continuous feedback loop which constantly adjusts sensor trigger and other parameters, ShotSpotter is able to deliver instantaneous system reports to dispatchers within seconds of a weapon being fired.

Seems like a good thing, but I wonder if the sensors can be turned on for other listening purposes. The low density of sensors might be one key factor that limits this type of use today.

My guess is that with only 8-12 sensors needed per square mile, a sound could have to be very loud to be noticed. Some articles say it only has an 80-ft accuracy. This could be by design, but probably has more to do with cost savings and 12 is apparently still sufficient to hear gunshots inside homes. Besides, more sensors could always be deployed. I just imagine someone will eventually want to tune the system to listen for certain words like “bomb”, gang slang, or drug terms.

Nonetheless, unlike cameras, which are criticized widely for full-scale surveillance in their typical setup, a sound-based system has the advantage of being tunable for known-bad activity. In that sense, it is easy to see how it grew out of earthquake sensing.

Coupled with cameras, sound sensors could in theory allow cameras to use the same known-bad activation, bringing automation and reducing privacy concerns of cameras. When a shot is fired, it would then spin up cameras and start recording in a specific direction. Even more into the future, imagine drones and/or robots that spring to life when they hear a sound and rush to a scene to start recording video.

Back to current issues, KCBS reportsa “successful” use of the technology in San Francisco:

“Even though the young man ran after the shooting, meaning he wasn’t at the location where the shooting occurred, the shot spotter technology pinpointed exactly where that shooting occurred. In this case it worked perfectly, exactly how it’s intended to work,” said Mannina.

The technology also allowed officers to secure physical evidence that they otherwise would not have found, had the new technology not been up and running.

ShotSpotter was used again early Monday morning when a gunshot was fired inside a house on Ceres Street. Several people were inside the house at the time, including a couple of small children, but no one was hurt. Police did arrest one person inside the house.

I like how officers are coupling technology with existing forensics processes to increase the accuracy of their investigations, but it begs the question of the accuracy of their information. I am sure it has already gone to trial but I wonder how people have argued the accuracy of the sound triangulation system?

Oakland is reported to have spent almost $400K on 84 sensors ($4,620/sensor) in 2006, and as a result was faced with a data analysis problem:

In its first year, the system detected nearly 3,000 gunshots, overwhelming city dispatchers. To counter the deluge, the police department worked with ShotSpotter to develop a mobile system through which officers would monitor alerts through laptops in patrol cars.

Interesting solution. Officers were equipped with more localized data, rather than having it route through a central dispatch system. Again, this could reduce privacy issues if local officers have to tie sounds to a case and thus no long-term central storage system is maintained.

It does not surprise me much that success of the system is said to depend on the talent, availability and training of those tasked with using it. This is just like any security logging and event monitoring technology:

ShotSpotter has proven more effective in some cities than in others. In North Charleston, S.C., for example, city officials say it helped to reduce the number of violent crimes in some of the more-dangerous neighborhoods by 35 percent in 2004.

But some larger cities have faced more limitations. Some Oakland lawmakers say that ShotSpotter has resulted in fewer than a dozen arrests since it was installed. The problem, they say, is that police don’t dedicate enough resources to follow up on the shooting calls.

Oh, well, I guess there still is no silver bullet solution. It is still a very interesting technology to watch, and I think it does far better integrated into other processes, rather than trying to stand on its own.