My aunt became permanently disabled after a car accident because my uncle was driving drowsy at that time. According to a national poll, over one-third of American drivers have admitted to driving when drowsy. Anti-Snoozer is a project powered by Intel Edison and Intel RealSense that is geared to prevent accidents that occur as a result of falling asleep behind the wheel. We believe reducing drowsiness would greatly improve transportation in Detroit.

What it does

Through applying RealSense’s advanced facial recognition technology, we’ve programmed the RealSense camera to detect four distinct characteristics of drowsiness: rapid blinking, closed eyes, yawning, and shifting of the pupils away from the direction of the road. The camera’s detection of these facial movements will concurrently employ Intel Edison to send out signals that are intended to alert and attempt to awaken the driver. These signals take action in the form of a buzzer, a wearable vibrator and crystal LED, the last of which emits red LED lights that both visually alert the sleepy driver to wake up and remain focused on the road, as well as warn surrounding drivers of an app-identified drowsy driver in their vicinity. The hope is that Anti-Snoozer will substantially reduce the number of sleep-provoked car accidents and generally promote safe driving.

How I built it

Using Intel RealSense SDK to do the facial and hand tracking, once drowsiness is detected, we send the data to Intel Edison for the beeping, vibration motor, and LED for warning. Additionally, if the phone is connected, we'd be able to vibrate the phone as well.

Challenges I ran into

Getting the algorithm to detect drowsiness, calibration for the algorithm.

Accomplishments that I'm proud of

Former Secretary of State John Kerry and Vice Premier of China had a chance to take a look at early version of the demo

What I learned

I've learned how to use Intel RealSense effectively and integrating it into IoT devices.

What's next for AntiSnoozer

The project is already open sourced, we are trying to upgrade the algorithm more accurately, and moving it using a Intel Joule so it can be built as part of independent IoT device

Share this project: