Inspiration
We're always losing our things and having trouble remembering where we left it behind. Our first thought is usual to retrace our steps but, more often than not, we can't remember these steps in the first place. With Backtrack, we wanted to reduce this feeling of being unable to remember and help individuals recover their lost items quickly.
What it does
Backtrack is a wearable device (i.e. watch, ring) that tracks your hand movements and position over time to detect when an item might have been placed down, judging rapid changes in acceleration, velocity, and position as the motion of putting something down. This data and, in particular, the points of interest outlined above are saved in a positional map and readily available whenever the user needs to find a misplaced item.
How we built it
We used a sensor with an accelerometer, gyroscope, and magnetometer connected to an Arduino to first take in the appropriate readings. From there, we performed a series of mathematical and physical calculations to fuse, filter, and process the data from the individual sensors with the goal of extracting the X, Y, Z positions and converting it to something representative of motion in 3D space. We then attempted to plot this data in real-time with MATLAB as we moved the sensor.
Challenges we ran into
The sensors were unfortunately not as sensitive as we wanted them to be so there was a decent amount of drift when calculating the position values. There was also lag in our readings that we had to account for because of its impact on real-time readings. We also wanted to integrate other tech (e.g. Myo Gesture Control Armband, Blynk) and processes (e.g. Wi-Fi communication, GPS integration, etc.) to bolster our system, but unfortunately these didn't end up panning out due to time and technical constraints.
Accomplishments that we're proud of
We're proud that we were able to learn the necessary Arduino and MATLAB skills over the course of the competition. We were also happy with how we minimized drift and adapted our work from 6DOF to 9DOF when the opportunity to use a better sensor presented itself later in the hackathon.
What we learned
We learned firsthand how challenging it is to integrate hardware and software, map data to 3D space correctly, and make sense of such data in that context.
What's next for Backtrack
Moving forward we'd like to revisit our code base to improve positional acquisition, build out plotting functionality to better visualize the results, store data in a database that enables efficient retrieval, implement "anomaly" detection to judge when items might have been place down, and improve our hardware tech stack as needed
Log in or sign up for Devpost to join the conversation.