Researchers at Purdue University Northwest have used AI to create a new indoor gunshot detection system that was specifically designed to overcome the current challenges of such systems.
The technologies have the potential to improve safety on school campuses and other public areas by enhancing situational awareness and reducing law enforcement and public safety officers’ response time to incidents involving gun violence.
According to project leader Wei Dai, assistant professor of computer science and director of the Advanced Intelligence Software Lab at Purdue University Northwest, traditional indoor gunshot detection systems have four drawbacks:
- minimal privacy protection due to using cameras for visual detection
- a high number of false alarms
- limited or no self-calibration
- and low affordability.
To address these challenges, Dai and team developed three different systems: a physical sensor, an AI-enhanced deep-learning model, and an algorithm for self-calibration and self-testing.
Sensor for air particles
The new physical technology utilizes air quality sensors to detect an increase in air particles following a gunshot, as shockwaves from ballistic impacts and muzzle blasts create pressure waves in the air that can be measured. During experimental testing, the sensor detected an increase in the number of 10-micron particles after a bullet was fired from 141 feet.
“It is a novel idea to detect gunshots without any concerns of privacy. Users can install air quality sensors in any room. They could monitor the air quality of buildings as well as the shockwave of gunshots,” said Dai. “Law enforcement and private security firms could also use it to detect gunshots as well as unexpected, dangerous air particles or chemical air in privacy-sensitive rooms and buildings.”
The researchers say the technology can be combined with other gunshot detection systems, including acoustic sensors, camera sensors and infrared sensors, to reduce false alarms.
AI for accuracy
The next technology uses AI to combine gunshot detection and fire alarm solutions. It identifies gunshots with high accuracy by using new deep-learning models trained with microphone sensors and air quality sensors. It runs on server-based algorithms and edge computers, which have the potential to expand the system to other compatible devices through an IoT system.
Self-evaluating algorithms
The third technology is self-evaluating algorithms for calibrating gunshot detection systems.
Traditional self-testing solutions are evaluated on acoustic gunshot sounds through powerful speakers or real gunshot tests. However, that can cause panic or unexpected mental pressure when tests are conducted during business hours. The new technology uses audio-coding algorithms for testing one or multiple acoustic sensors.
“This self-evaluating innovation is encoding encrypted audio sounds,” said Dai. “It will not impact most people and regular business operations. It allows users to evaluate acoustic gunshot sensors at any time and in any location.”
During experimental testing, Dai’s team tested the audio coding algorithms in classroom hallways during business hours for two weeks. He said most students and faculty did not notice the tests.
No false alarms
The Purdue University Northwest research team has been installing and testing sensors in university facilities since January 2024. The AI-driven gunshot detection system has collected 3,631 sounds similar to gunshots—with no false alarms.
In fact, there were no false alarms even after testing the system with more than 220,000 sound simulations like nail guns, hammers, thunder, firecrackers, backfiring cars, helicopters, popping balloons and blank ammunition.
The system can send alarm messages to police officers via mobile app or website.
Now, the team is moving on to conducting large-scale tests.
“We are partnering with several American universities, including Indiana State University, to evaluate the gunshot detection technologies for campus use, and looking for more to join the large-scale tests,” said Dai. “Scientists in Greece and Japan believe that the technologies could detect gunshots or explosions at subway stations.”
Dai, via the Purdue Innovates Office of Technology Commercialization, has applied for patents on all three of these new technologies. He is looking for industry partners interested in developing or commercializing the technology.