Gunshot Detector

Inspiration

We would like to be able to assist law enforcement in responding to gunshots detected in the Philadelphia area. By notifying officers immediately, we can decrease response time and improve safety in the city.

What it does

We train the detector with a sample gunshot sound, and then the detector continuously listens for loud sounds and processes the audio to determine if the sound was a gunshot or not. If it was, it connects to the LoRaWan network and informs a user that a gunshot occurred.

How we built it

We used an mbed platform (mDot) with LoRaWan technology. All programming was done in C++. We connected a microphone to an analog input pin of the mbed, continually sampled the microphone reading and stored the values in an array to be processed. We then turned these samples into the frequency domain and analyzed the probability that it was a gunshot.

Challenges I ran into

Digital signals processing was difficult to implement on a microcontroller. Additionally, while attempting to store long audio signals, we ran out of memory quickly. Finally, connecting to the LoRaWan network, which is a new technology, took a lot of time given the lack of documentation.

What’s Next

We would like to add a GPS on board for location/accurate time information. Additionally we would like to make our DSP algorithm more robust. Finally, we would like to use a better quality microphone/amp/filter setup.

Built With

• c++
• mbed