"What the eyes see and the ears hear, the mind believes."
Harry Houdini

Keeping an Ear on Crime

The NSA is listening to our phone calls. The FBI is using face detection to catch wanted criminals. Apu at the Kwik-E-Mart is watching Bart Simpson with surveillance video. And now the police are listening for gunshots in neighborhoods across the nation.

Like GPS, radar, and the microwave oven, technology developed for the battlefield has found itself on Main Street. Gunshot detection is another military trickle-down technology that police are using to protect our citizens. Police departments all over the world are placing these listening devices in urban areas that have a history of or potential for high crime rates. Most systems detect, analyze, and alert police within five seconds of a suspected gunshot.

In the mid 1970s, the Navy started development on devices that they hoped would locate artillery fire and track bullets in flight. Other agencies had similar attempts over the next few decades with limited results. But the Bosnian War of the early-to-mid 1990's precipitated a flurry of new development.

Bosnian paramilitary snipers were perched on top of ridges and buildings in and around the capital city of Sarajevo and taking pot shots at everyone. When the United Nations finally stepped in, they not only needed ways to monitor the battlefield and city for snipers, but a way to ensure a cease fire would hold up. Several companies provided acoustic-based solutions.

The basic solution was to listen. With an array of microphones placed around an area of interest, detectors would listen for loud, nearly instant sounds that were similar to gunshots from sniper rifles. Early success led other governments around the world to step up their research into "sniper" detection.

Here in the states, sniper detection research was already under way by the US Army under the umbrella of the Defense Advanced Research Projects Agency (DARPA). The DARPA initiative, not to be confused with the Dharma Initiative from Lost, launched many new approaches to sniper detection, such as infrared detection of heat from weapons. Another new method was the use of digital audio analysis of the analog signal. Nearly twenty years later, the Bullet Ears acoustic counter sniper system developed by BBN Technologies is still considered state-of-the-art.

Another technology that came out of the DARPA initiative was one developed specifically for urban environments, the System for Effective Control of URban Environment Security (SECURES) system. SECURES was first used in Olympic Village at the 1996 Atlanta Olympics. ShotSpotter and SENTRI are similar urban gunshot detection products from private companies introduced in the last couple of decades.

So how does the system work? In general, it listens for any sound above a certain loudness threshold. The SECURE system for example, analyzes the sound in question from an array of microphones by first determining it's length. If it's longer than 0.6 milliseconds (< 0.6 ms), it assumes it is either a .22 gun with blanks, a door slam, a hammer, a bell, or a balloon (that's some party going on!).

If it is longer than 0.6 ms, then a multitude of calculations occur. The different products vary from one another in this phase, but all try to do the same thing - make sure it's really a gunshot. One tries to distinguish reflections from surfaces. Another tries to discern between a blast and impact using harmonics. The narrowness of the sound pattern is also scrutinized. Gunshots are unidirectional in their sound pattern, but fireworks for example, are more omnidirectional.

To find the location of the shooter, these systems use hyperbolic multilateration. Huh? Well in basic terms, we have two ears for a reason - to figure out where a noise is coming from. If a bear roars on your left, the sound hits your left ear first. A millisecond later it hits your right ear. Your brain tells your head to swivel to the left. It then tells your legs to RUN! to the right. The gunshot detector has four microphones for more accuracy.

Current gunshot detection systems are far from perfect, but research in the labs now will soon show up in the streets and parks of America. So go ahead and have that birthday party in the park. Don't worry that the police will crash your party when the birthday girl pops her balloon. (I hope the system knows what a piñata being hit sounds like!).

Did You Know?

• The accuracy of finding the shooter's location is from 25 to 75 feet in most systems.
• How accurate are the gunshot detectors? ShotSpotter averages 67% accuracy in true positive activations.
• Studies show a great reduction in response time to shootings using these systems versus a 911 call alone.
• A typical police handgun's loudness has been measured at 140dB-SPL at close range, just 10 dB softer the sound of a jet engine 3 feet away.
• Each gun can sound different. For instance, a revolver (think Old West cowboy) has a precursory blast at the rear of the barrel from the ignition before the main blast from the front of the barrel. A pistol only has one main blast.
• The National Park Service has begun installing gunshot detection systems to control illegal hunting of protected animal species.
• ShotSpotter, Inc. acquired the SECURE product line in 2009, becoming the dominate market leader.
• In 2014, 64 cities in the U.S., plus Canada, Puerto Rico, Mexico, and Brazil have ShotSpotter systems installed.

Tech Notes

Most gunshot detection systems can determine what kind of weapon was fired, but not with complete accuracy yet. Shotguns are the easiest, followed by pistols. But automatic rifles are very difficult. The system compares pattern data from the gunshot with a database of parameters of different models. Temporal and direction characteristics are part of those parameters stored in the database.

To figure out the acoustical signature of a gun, scientists can study two processes that produce sound. The release of the projectile from the barrel emits a piston-like sound, hence, an acoustical signature that can be measured. If the projectile is traveling faster than the speed of sound, as in a rifle, there is a ballistic shock wave (a Mach cone). This shock wave actually follows the projectile - a miniature sonic boom. But most handguns, like those used by delinquents in crime, fire bullets at subsonic speeds, so there is no "sonic boom," just a muzzle blast wave. All current gunshot detection systems are programmed to analyze muzzle blast waves, which have completely different acoustical properties than shock waves.

Curiously, forensic audio scientists can determine which way the barrel of gun was pointed when analyzing monophonic telephone and video camera recordings, but current gunshot detections systems don't employ this method.

Neil Kesterson

"Gunshot Detection Systems in Civilian Law Enforcement", JUAN R. AGUILAR (Laboratorio de Electro Acustica & Procesamiento de de de Senal, Santiago, Chile)
Journal of the Audio Engineering Society ENGINEERING REPORTS Vol. 63, No. 4, April 2015 (C 2015)