TrippinTracker | Devpost

Inspiration

We were inspired when brainstorming problems facing high schoolers in our country. We realized that binge drinking was a problem that plagued students across the nation. Our product will allow concerned friends to remotely monitor and look after someone at risk of getting drunk.

What it does

Our product tracks footsteps and uses it to classify an individual's walking pattern as regular or drunk. It then communicates via the app to notify if drunk walking is detected.

How we built it

We used an Arduino connected to two ultrasonic sensors to detect steps. We then pass this data into a GRU Neural Network to classify the sequence over 10 seconds of walking as regular vs. inebriated.

Challenges we ran into

A major challenge we encountered was connecting the ultrasonic sensors to the Arduino. Since we needed data from both sensors simultaneously, we had to connect a sequence of wires from each sensor. Each sensor was located at the individual's ankle, so the wires had to reach the upper leg to connect both. Our secondary challenge was communicating between the PC and the Arduino and using the Neural Network in real time. We circumvented this by using special python libraries to communicate directly with Arduino.

Accomplishments that we're proud of

An accomplishment that we're proud of is how we collected the data. We attached the Arduino and sensors in a makeshift way to our teammate. We then had him walk to collect over 500 data samples for each regular and drunk walk.

What we learned

Before this hackathon, we didn't know how LSTMs worked. However, over the course of this hackathon, we were able to learn about Neural Networks that process sequential data to produce a meaningful output.

What's next for TrippinTracker

The following steps for TrippinTracker are making it more fine-tuned and more robust. In terms of software, we need more data so that we can accurately predict minor levels of alcohol consumption. We can also use accelerometers and magnetometers to detect a position and swaying, incorporating that into our data. In terms of hardware, wireless communication between the sensors and the Arduino would vastly improve usability. Finding a way to combine it with day-to-day clothing effectively would make it more practical.