Prototype & User Testing

    For the first prototype, I used an old 20-inch LCD as the screen. I ordered a piece of plexiglass with the mirror film pre-applied to it, and I taped it onto the display with masking tape. The masking tape on the edge simulates the frame, and it worked pretty well. I used a Kinect contour tracking code for this version. By finding the highest point in the image, which most of the time is going to be the head, I could achieve the effect I wanted with the contour line of the audience in front of the sensor. I took this project to the MIT media lab for user testing, and I received affirmative feedback. People had similar thoughts to when I showed them the original concept photos, and everyone had a different definition for the light. This was a successful prototype, which established the system for me. However, it wasn't perfect. Due to the contour data I was using, it was merely tracking the highest point in the data and pinning my image to that location. The computer wasn't able to distinguish a person from a group of people, which resulted in the light jumping around when there were two users in the frame at the same time or when someone raised their hand higher than their head. Therefore, I switched the software to a MAX plugin called dp.kinect which recognizes people as joints and skeletons. It tracks in real-time (without needing the awkward Kinect setup pose) and is very efficient. Because it uses the skeleton of people, I can always get the position of people's heads–this means that it won't be affected by people raising their hands higher than their heads or multiple people being in the frame at the same time.