(Photo by Tsvetoslav Hristov on Unsplash)

BOCA RATON, Fla. — Mass shootings have many Americans constantly on edge when they’re out in public. With nearly 300 occurring in 2021 alone, being able to tell the difference between gunfire and a harmless plastic bag popping could mean the difference between life and death in a crisis. When our ears fail us, however, a new study finds high-tech sound equipment can learn to spot real gunfire and cut down on false alarms.

Despite rising gun violence in many cities, there is still some hesitation to install acoustic gunshot detector systems in certain areas because of how expensive and unreliable older models can be. A team from Florida Atlantic University’s College of Engineering and Computer Science tried to fix these issues and lower the false-positive rates of these devices.

3 in 4 gunfire detectors report false-positives

outdoor park
Researchers recorded gunshot-like sounds in locations where there was a likelihood of guns being fired, which included an outdoor park. (Credit: Florida Atlantic University)

To do this, researchers created a dataset containing audio recordings of plastic bag popping in various locations. The sounds ranged from 400 to 600 milliseconds in length and featured bags of all sizes “exploding” for the microphones.

The team then created a computer algorithm based on a convolutional neural network (CNN) to sort out the various sounds coming from the bags. Adding in a database of gunshot sounds, study authors trained their thinking computer to weed out which sounds were actually life-threatening gunshots, and which were harmless street noise.

Results show that fake gunshot sounds (or similar startling noises) easily confuse an untrained sound detection system. In fact, standard gunshot detectors misclassified a staggering 75 percent of plastic bag pops as gunfire. Even a deep learning model trained to sift through common urban sounds couldn’t tell the difference.

However, when the team added in their plastic bag training into the model, the CNN classification system could accurately pick out dangerous gunfire from urban noise.

“As humans, we use additional sensory inputs and past experiences to identify sounds. Computers, on the other hand, are trained to decipher information that is often irrelevant or imperceptible to human ears,” says Dr. Hanqi Zhuang, professor and chair of FAU’s Department of Electrical Engineering and Computer Science, College of Engineering, and Computer Science, in a university release.

“Similar to how bats swoop around objects as they transmit high-pitched sound waves that will bounce back to them at different time intervals, we used different environments to give the machine learning algorithm a better perception sense of the differentiation of the closely related sounds.”

More data leads to smarter detectors

The researchers recorded many gunshot-like sounds in eight indoor and outdoor places where someone would likely hear a gunshot. They started with various kinds of bags, including one that created the most gunshot-like sound — trash can liners.

To find out which of these sounds confuse a gunshot detector most often, researchers first trained their computer model without letting it hear plastic bag pops. This included teaching the computer 374 different varieties of gunshot sounds coming from an urban sound database. Aside from plastic popping, these common urban noises range from dogs barking to car horns, to sirens, jackhammers, and children playing.

“The high percentage of misclassification indicates that it is very difficult for a classification model to discern gunshot-like sounds such as those from plastic bag pop sounds, and real gunshot sounds,” explains first author Rajesh Baliram Singh. “This warrants the process of developing a dataset containing sounds that are similar to real gunshot sounds.”

Study authors conclude that adding even more diversity to the range of noises which scientists use to train these devices will only benefit cities looking to cut down on gun crimes as well as false alarms called into the police.

“Improving the performance of a gunshot detection algorithm, in particular, to reduce its false positive rate, will reduce the chances of treating innocuous audio trigger events as perilous audio events involving firearms,” says Dr. Stella Batalama, dean of the College of Engineering and Computers Science.

“This dataset developed by our researchers, along with the classification model they trained for gunshot and gunshot-like sounds is an important step leading to much fewer false positives and in improving overall public safety by deploying critical personnel only when necessary.”

The study is published in the journal Sensors.

About Chris Melore

Chris Melore has been a writer, researcher, editor, and producer in the New York-area since 2006. He won a local Emmy award for his work in sports television in 2011.

Our Editorial Process

StudyFinds publishes digestible, agenda-free, transparent research summaries that are intended to inform the reader as well as stir civil, educated debate. We do not agree nor disagree with any of the studies we post, rather, we encourage our readers to debate the veracity of the findings themselves. All articles published on StudyFinds are vetted by our editors prior to publication and include links back to the source or corresponding journal article, if possible.

Our Editorial Team

Steve Fink


Chris Melore


Sophia Naughton

Associate Editor