Penn criminologist demonstrates wearable gunshot detection

Community-based offenders, such as those on house arrest or out on bail or parole, account for a disproportionately large share of the perpetrators of fatal and non-fatal shootings.

Gun Detection
Deputy Chief Michael Fink of the Penn Police Department participates in the study’s gun range data collection. Photo by Addie Métivier

Community-based offenders, such as those on house arrest or out on bail or parole, account for a disproportionately large share of the perpetrators of fatal and non-fatal shootings. Detecting and deterring shootings among this population can be challenging in the absence of reliable evidence that a particular community-supervised offender illegally used a firearm.

New research from Penn’s Department of Criminology has the potential to transform the national system of corrections by tracking gun use by community-based offenders.

In a study published in the journal PLOS ONE, Charles Loeffler, the Jerry Lee Assistant Professor of Criminology in the School of Arts & Sciences, demonstrated the feasibility of using low-cost, wearable inertial sensors to detect firearm usage.

To conduct the study, Loeffler used sensors similar to those found in fitness trackers to recognize wrist movements and other signals corresponding to firearm use. Research participants were divided into three groups with each participant wearing a tri-axis accelerometer watch on his or her right wrist.

Gun Detection
Deputy Chief Michael Fink of the Penn Police Department participates in the study’s gun range data collection. Photo by Addie Métivier

The first group included 10 officers from the Penn Police Department who were asked to participate in a shooting task with six handguns that ranged from a .22 caliber to a .45 caliber weapon.

The second group consisted of three members of the general public who were instructed to engage in their normal daily life routines from morning until evening, and wore the sensors for six to eight hours at a time. 

Five construction workers comprised the third group; the workers were directed to engage in normal construction and demolition tasks, including using pneumatic nail guns, pneumatic jack-hammers, .22 caliber powder-actuated fastener guns, and other construction tools.

The wearable sensor data was used to train a detection algorithm that achieved more than 99 percent accuracy in classifying individual gunshots, demonstrating that firearm use can be reliably distinguished from a range of potentially confusable human activities.

It turns out that gunshots are highly distinctive events when viewed from the perspective of the human wrist, Loeffler says.

“The wrist experiences a near-instantaneous blast wave that is closely followed by the recoil impulse,” he says. “The entire event is over in a fraction of a second.”

According to Loeffler, whose research focuses on the effects of criminal justice processes on people’s lives, the low-cost wearable sensor technology offers criminal-justice practitioners a potential alternative to the existing monitoring systems that are not specifically designed to detect individual firearm usage.

“If integrated sensibly into existing community-supervision systems, it could enhance the ability of correctional authorities to deter and or detect firearm use while allowing community-supervised populations to experience less onerous conditions of release,” he says.

Originally published on .