December 8, 2005 By Jim McKay, Justice and Public Safety Editor
So what if the gunshot automatically triggered a 911 call, and captured video of the shooter? Police in Chicago are hoping to curb gun violence with technology that does just that.
The technology -- Smart Sensor Enabled Neural Threat Recognition and Identification (SENTRI) -- recognizes the sound of a gunshot within a two-block radius, pinpoints the location of the shot with a surveillance camera, focuses on the location, and in less than 1 second, places a 911 call.
The goal is to use the devices to prevent homicides in areas known for gang activity and gun violence.
Chicago successfully deployed 53 surveillance cameras over the years, and has deployed the gunshot-recognition technology in about one-third of those. The cameras, by themselves, were credited with reducing the city's 2004 crime rate to its lowest level since 1965 -- sexual assault is down 5 percent from the previous year, robbery is down 8 percent, aggravated assault is down 5 percent, and total violent crime is down 7 percent -- and it is hoped the SENTRI system will provide even more ammunition against crime.
"The goal is to integrate our surveillance with this gunshot technology and to be able to send these gunshot alerts to our crime detection specialists that work in our observation center [located in the 911 center] every time we receive an alert," said Chicago Police Sgt. Greg Hoffman. "The camera will point in the direction of that gunshot [and] report immediately; the officers here in the detection center will have the ability to communicate to first responders about actionable intelligence about what they see."
The units were installed this spring, but Hoffman said it's still too early to cite any tangible results, and that the implementation is still in the pilot stage.
"Pilot stage is really a technical term," said Bryan Baker, chief executive officer of Safety Dynamics, which produces the gun recognition technology. "When they define something as a pilot, that means there's still a certain confidentiality about information. Once it gets reclassified as production and not pilot, then all the information becomes nonconfidential."
The Los Angeles County Sheriff's Department also was installing 20 of the units late this year, and San Diego and San Francisco police were preparing to launch test projects. Police departments in Philadelphia, New Orleans and Atlanta also expressed interest in the technology.
The units, basically a microphone surveillance system, are located 20 feet above the ground and usually at right angles of buildings to provide coverage on each side of a building.
The acoustic sensor recognizes a sound as a gunshot and sends range and bearing data to the camera, which starts out with a wide angle, then tilts and zooms in on the area. Then it sends a 911 message to the 911 call center, according to Baker.
"Most of the time the shooters shoot and then run, so by coming in at a wide angle, if they catch somebody actually running, they'll at least have that on camera as it zooms in," Baker said. "The command center will take control of the camera at that point, and go wherever they want to go with it."
Like the Human Brain
The technology was developed by University of Southern California professors Ted Berger and Jim-Shih Liaw, who delved into brain cell research to create a technology that recognizes the digital markings of a gunshot sound while excluding most other sounds.
"There's an algorithm modeled after the way the brain works," Baker said. "Much like the human brain, where we recognize sound by hearing it and somebody else telling us what it is or us visually seeing what it is, we then associate that sound to a particular thing, like a gunshot."
In developing the technology, the professors recorded gunshots under different conditions and through obstacles, such as rain, wind and trees. They used these recordings to identify a common digital pattern -- the pattern of the sound of a gunshot.
The device is programmed to recognize the digital pattern of the gunshot sound and ignore all other sounds. "It's able to pick out that pattern even in the midst of other noise," Baker said. "Even when there's noise louder than the gunshot itself."
There is the possibility of false positives, especially with loud firecrackers, Baker said, and there are other obstacles to performance, such as buildings and trees. "Video is even less useful [in those conditions] because you have no line of sight," Baker said. Still, the device will recognize that there was a gunshot in the general area and notify police, although the exact location might not be available.
That's why the devices are placed on the corners of buildings -- so that two devices placed on opposite corners of a building would cover an entire block.
Theoretically a city could be saturated with the devices, but that's not practical with a cost of $4,000 to $9,000 per unit. The units are available through several integrators.
Safety Dynamics supplies the hardware platform and software to the integrator, which is responsible for producing the video camera and the wireless or fiber-optic communication mechanism by which the data is transmitted, and for developing the overall box or outer shell that houses the technology.
Engineers are improving the same technology to provide perimeter defense-type mechanisms. The same principle is applied, but instead of gunshots, the mechanism would recognize the sound of someone cutting or climbing a chain link fence.
Several companies are creating a video analysis component that would recognize a shooting scene -- the position someone would be in when holding a gun, someone lying on the ground or a group of people running.
"They can digitize what somebody holding a gun would look like," Baker said. "Then they can lock in and follow him. When you put all these things together, it can be successful in protecting parameters. You make it increasingly difficult for an intruder to escape notice."
In effect, it's technology with eyes and ears, and the functionality to use them. "It proactively tells the camera there's something happening over here," Baker said. "Go over there and look at it and send all that data back."
You may use or reference this story with attribution and a link to