We spent a couple of hours deciding on a project to pursue. We had a 15 minute silent ideation session where we put ideas on post-its and then discussed individual learning goals. We were able to weed out a few ideas from an incompatibility with learning goals. Through discussion, we narrowed it down to five main ideas (left picture) and discarded all the rest (right picture). We finally decided on a page turner for musicians. We also formulated our MVP: "We will design and fabricate a page turner for musicians that is actuated with a hands-free method."
Prototyping mechanical design
At this point, we were unsure of which type of sheet music to target: a stack or a binder of sheet music. We developed a design with two wheels and a swiper mechanism, as this allowed us to test both options. For the stack of paper, the right wheel will turn and push the paper toward the center. Then, the left wheel will continue to move it out of the way.
In binder mode, the right wheel is used along with a swiper. When the wheel moves, it creates a pocket with the paper, and the swiper finishing turning the paper. Below is a sketch model of our system, created mainly with cardboard. This illustrates the approximate scope and positioning of the various components.
To ensure a fine degree of turning accuracy for our wheels, we decided to use stepper motors. By the end of this sprint, our system moved one of the wheels via a button input.
Software: Eye Tracking
Our proposed method for triggering the hands-free turning mechanism was some form of eye tracking. Ideas for a "trigger" included looking at a specific place, blinking in a certain pattern, and winking. We used OpenCV to implement an initial program that detected eyes open and closed. While we did get the detection working, the system was not very reliable. It heavily depended on the lighting and the person in front of the camera.
Our Main Takeaways
After testing the blink detection software, we found it to be too unreliable to be sensible for this application. In particular, it worked on less than half of the participants tested and it required a particular type of lighting to be accurate. We decided to pivot to beat detection in the next sprint.
During the sprint, we ran into issues surrounding our stepper motors. In particular, the built-in Arduino stepper motor library was unable to run both stepper motors at once. This made it difficult to test moving a stack of sheet music. Moving forward, we will be focusing on moving sheet music in a binder format. Not only will this eliminate the need for a second wheel, but it allows for more flexibility in regard to future expansion. In particular, the design can be expanded for bound books of sheet music, and a reverse function can implemented without a full redesign.
The PowerPoint that we presented in class can be found here: