Here’s How the NYPD’s Expanding ShotSpotter System Works
May 18, 2016
BROOKLYN — Last year, the city’s police department started using ShotSpotter, a technology that helps police automatically detect the sound of gunshots.
In a pilot program started last spring, the company deployed hundreds of tiny sensors on rooftops and light posts in several gun violence-prone precincts in Brooklyn and The Bronx.
Now, the city is looking to significantly expand the ShotSpotter technology, rolling it out to all five boroughs with a $3 million allocation in the mayor’s proposed budget.
The funds are pending approval by the City Council, which must make its decision by June 1.
The equipment picks up gunshot blasts, analyzing the data and relaying that information to police, all within between 30 to 45 seconds after a shooting, according to ShotSpotter CEO Ralph A. Clark.
Throughout the year, ShotSpotter started showing up in police reports and shooting investigations; in April, the system went live in Upper Manhattan and more areas of the Bronx and Brooklyn, the NYPD said, covering a total of 24 square miles.
To find out how exactly the technology works, who has access to the data and where ShotSpotter hopes to deploy in the future, DNAinfo spoke with Clark about the company, founded in the mid-1990s with its first gunshot detection system installed in Redwood City, Calif. in 1997.
The interview has been edited for length and clarity:
DNAinfo: How does ShotSpotter detect a gunshot? How does it work and what technology and equipment is involved in the system?
Ralph A. Clark, ShotSpotter: The first thing starts with the sensors — it’s essentially a computer with microphones on it and a lot of software. It has a chip on it and what it’s designed to do is basically to ignore ambient noise and trigger and timestamp on impulsive noises — booms or bangs. And when we deploy the system, we’ll build an array out — a network of sensors — and there will typically be 15 to 20 sensors per square mile, although in the case of New York City, because of the acoustic environment with buildings and noise, I think we’re [at] more like 20 or as many as 25 sensors per square mile in certain locations.
We try to space the sensors out in a way that at least three sensors will be able to hear that boom or bang noise. It ignores everything else, but when it hears a boom or bang, it will timestamp. It will say “Yup, I heard a boom or bang” and it’ll timestamp it down to the millisecond.
The most important piece of technology we have in the sensor is a GPS chip [that] gives us a precise location and precise time so when these sensors detect the boom or bang and timestamp it, they’re sending that little bit of metadata … back to the software in the cloud [that triangulates the location]. It’s like solving a puzzle, right? It says “OK, the location of that boom or bang sound had to be here.”
Then it will send the alert to our [California] headquarters location where we do our 24/7 monitoring review service where we have trained acoustic experts make the final determination.
[We] will have a three or more sensors worth of audio and that’s what our acoustic experts are listening to — that snippet of audio that the sensor heard.
A number of times, they’ll listen to it and say ‘You know what, that’s not a gunshot’ and they’ll dismiss it. And that’s the way we reduce false positives. We don’t want to alert an agency like the NYPD, or any agency for that matter, [for] false alarms. We really have these people carefully listen. And then when it passes their muster, they basically push a button and that digital alert shows up in a 911 call dispatch center and can also show up real time on a person’s laptop computer or even their mobile phone. And all this takes place in about 30 to 45 seconds.
But I would strongly suggest that the real magic happens once an agency gets the alert. What do they do with that? Technology by itself can’t defeat gun violence. It requires a much more deliberate and broad strategy. We’re kind of the awareness strategy … that make police aware of all the gunfire incidents that are happening because we know from our experiences that a full 80 to 90 percent of the time, believe it or not, when gunfire is played out in underserved communities, people don’t call 911.
[Editor’s note: According to the most recent figures from NYPD, between 75 and 80 percent of shooting incidents are not reported to 911 in the city.]
What kinds of gunfire does ShotSpotter have difficulty detecting? Could the system detect, for example, a gun fired in an apartment building?
We certainly can’t detect indoor gunfire, although we have occasionally, but that's more accidental. It’s all around the acoustic energy that’s generated from a muzzle blast. So, if someone fires a gun indoors, there’s not enough acoustic energy to go out and reach the sensors that they can then hear the impulse noise and timestamp. And if people do execution-style shootings, there’s a possibility that we don’t catch those. We have some examples where people have literally put the barrel of the gun to someone’s body. The body absorbs the acoustic energy, so we’re not going to detect that.
As amazing as the technology is, and as fast as we are at striving for perfection, I will make the point that we are going to have missed detections — we call those false negatives. Those are gunshots that we should have detected [and] for some reason we did not. They tend to be very rare and we try to keep those to a minimum, of course.
Let’s talk some more about the audio recordings gathered from the sensors. Where do those recordings go? Does the NYPD have them?
We push that to the NYPD. We push that to the patrol officers.
They’re able to analyze it, as well?
Yes. It’s already analyzed … We’ll annotate our alerts with context. Our folks are so good at this stuff, they can tell if it’s a multi-shooter event. They’ll annotate the metadata side of the alert page to say if we’re dealing with a multi-shooter event or, 'Hey, this sounds like a semi-automatic.' I think it’s helpful for a police officer to have a little bit more auditory context about what they might be stepping into.
There are at least two instances — one in Oakland, Calif. and another in New Bedford, Mass. — of ShotSpotter recordings used as evidence in trials. How often does that happen?
Those are very rare. In those two particular cases, there was someone literally shouting over or just before or just after they got shot. The system clips [the audio recordings] off in the front, so we’ll cue it up and I think it’s one or two seconds before the gunshot event — boom, boom, boom, boom — the gunshot, and then it will play another two or three seconds after. You need that in order for our reviewers to do their work.
In the case of the Oakland situation, because the person, right after he got shot, he said ‘So and so, why’d you do me like that?’ and he yelled it out, so that was heard and picked up by our sensors. There’s nothing we can do about that. There’s no privacy issue at that point because it’s a public setting. If you’re shouting out when you get shot, that’s not presumed to be a private conversation. But I can tell you, that’s extremely rare. We’re essentially publishing 60,000 gunshots a year … and I think there’s been about four times where there’s been someone yelling over or on top of a gunshot clip.
Is there any scenario in which ShotSpotter would record conversations between people?
The short answer is no. We take the privacy question very seriously and we embrace that privacy question. You know, I live in Oakland, Calif. Berkeley, Calif. I grew up in the sixties. We don’t fear the question being asked. We were leaning in on this. We actually went directly to the ACLU [American Civil Liberties Union], visited them directly, because we wanted them to understand exactly how this technology works, what it does, what it doesn’t do, how it’s designed.
One Brooklyn reader of DNAinfo who lives in an area included in the ShotSpotter pilot program told us he is worried that the technology is a “dragnet” that will increase monitoring of high-crime areas, without necessarily reducing gun violence. What can you say to that person who is worried about mass surveillance of his neighborhood?
We are a surveillance technology. There’s no getting around that. But it’s a surveillance technology that’s completely passive and we’re only detecting when a felony is in commission. So the NYPD — and us, for that matter — is only getting alerts when a gun is fired or when a possible gun is fired. And when we figure out that it’s not a gun that’s being fired, from a human point of view, we dismiss that and that doesn’t even get sent to the NYPD. And that just allows a level of precision for policing that’s a game changer.
I think the way Mayor de Blasio puts this technology — I think it’s perfect. I think he stepped into the breach, which I’m personally happy about being an African-American male, and said, hey, we are going to eliminate broad stop-and-frisk, but what we’re going to do in replacement of that, we’re going to be much more precise in our response and only respond when we know something is going awry. And that’s where ShotSpotter plays a very significant role. We are interested in being deployed where, unfortunately, urban gun crime exists. There’s no value in us being deployed in a place where people don’t shoot guns and so, that’s where we go.
Can you say how ShotSpotter has reduced gun violence, shootings or increased arrests?
We’re part of the overall strategy. We’re not singular and solely responsible for anything. The true measure … is the actual reduction of gun violence. It’s not an arrest, it’s not homicide … but what we like to do is measure year-over-year gun violence in the same coverage area. And it just takes a while to do that. You’ve got to be deployed two years to do a year-over-year analysis. Where we’ve done that, we’ve seen agencies drive gun crime down 25, 35 percent.
What we’ve come to realize through our work with agencies, it’s very few criminals that drive the gun violence problem. These are called chronic shooters or serial shooters or trigger pullers. So what happens is, if you’re able to — through very rapid response to all shooters, you investigate shootings, you recover shell casings … that can help you more quickly identify who that serial shooter is and, in the process, you can deter them. And when you take a serial shooter out, not only do you take their shooting out, you take out retaliatory shooting out from the other side, as well, and the related crews.
So I don’t have any doubt when we go to New York City, just knowing how they’re resourced and how they’re managed, that we’ll see really good reductions, year-over-year.
Where do you envision the technology being used in the future? Where else does ShotSpotter hope to expand?
We’re deployed in about 90 cities. We think there’s probably a need in about 1,000 cities in the U.S. There’s a global need, as well, in Central America, parts of South America and parts of Africa, South Africa primarily. In fact, we’re going to be lighting up a system in Capetown pretty soon.
We’ve got a lot of work to do in being deployed in more cities, working with more agencies, helping agencies develop more effective gun violence abatement strategies. The technology is invented ... so we have the opportunity to spend much more of our time with outcomes — how you respond, how you investigate. We’re kind of moving up the food chain to that level. We consider ourselves, frankly, gun violence abatement consultants that happen to use the technology.