GuardianShot

Logo from website

*Here is my research paper, which provides more detailed documentation on my engineering process for this product! * link

Inspiration

The growing threat of gun violence in schools across the United States has underscored the urgent need for accessible and efficient safety measures. Technologies like ShotSpotter and Shooter Detection Systems are incredibly effective but often out of reach financially for many schools due to their high costs. This gap in accessibility spurred the development of GuardianShot, an affordable, real-time gunshot detection system designed specifically for the school environment, aiming to bolster security and provide peace of mind.

What it does

GuardianShot leverages a Raspberry Pi and a USB omnidirectional microphone to continuously monitor school environments for gunshots. Upon detecting a gunshot using its sophisticated Convolutional Neural Networks (CNNs), the system immediately sends an email notification to the relevant authorities, facilitating a quick and informed response to potential threats.

How we built it

My journey began with initial prototypes that used an Adafruit microcontroller and a MEMS microphone to display audio data in real time. As needs evolved, I upgraded to a Raspberry Pi, enabling us to handle complex machine-learning algorithms. Developed three distinct CNNs to analyze audio spectrograms, ensuring the system could differentiate between gunshots and other loud background noises accurately and reliably.

Challenges we ran into

One of the toughest challenges was achieving high accuracy in detecting gunshots amidst various background sounds, and significantly reducing false positives. More information regarding this topic is in the attached research paper I wrote. Transitioning from simpler microcontrollers to more robust mini-computers like the Raspberry Pi required me to scale my technical skills quickly and learn new difficult concepts. Additionally, refining the machine learning models to effectively identify gunshots proved to be both challenging and educational.

Accomplishments

I was able to successfully create a working prototype that achieves our engineering goals without the hefty price tag of existing gunshot detection solutions. Integrating CNNs that can accurately detect gunshots in real time was a significant accomplishment. More importantly, I am proud of making this crucial technology accessible to schools regardless of their budget, enhancing safety and security in educational environments.

What we learned

This project deepened my understanding of audio processing, Fast Fourier Transform, neural networks, and real-time data processing. I also learned about the challenges of implementing technology solutions in sensitive environments like schools. Striking a balance between technical goals and practical legislation constraints taught me valuable lessons in engineering and innovation.

What's next for GuardianShot

I plan to continue refining the accuracy of this system and revamp the dashboard. Field tests in various school and outdoor environments are up next, to collect more false positives and improve the system’s reliability. I am meeting with Chief Sult at an outdoor gun range in a couple weeks to get more reliable feedback with real gun sounds. I also hope to expand the system’s capabilities to integrate with existing security systems in schools and explore potential applications in other public spaces such as libraries and community centers.