Initial Idea for Project 3

Hello, Everyone,

We are posting this quick sketch of an idea to get initial feedback. The scenario below goes beyond the scope of what is possible within the next several weeks. From it, we extract what we think is a compelling scope of work. Your feedback is greatly appreciated.

Miss Jane wakes in the morning, feels around for her glasses, and pulls her walker up to the bedside. The television is in the middle of the room facing her. As she gets up and proceeds to the kitchenette, the television tracks her movement, follows her and orients itself to her position. When she turns and faces it, it recognizes that she is looking directly at it and waits for instruction. Simple standard gesturing allows her to turn it on or off or change the channel and volume. As Miss Jane progresses through her daily routine, the television passively tracks her position, follows her and waits for her to gaze at it. The television uses a combination of ultrasonic sensors and IR sensors to locate Miss Jane and determine gaze. This allows her to watch TV from anywhere in the apartment with a single television. Since Miss Jane follows a more or less set pattern in her daily routine, the television uses a neural network algorithm to anticipate expected behavior. This allows the television, after just a couple of days, to preposition itself based upon her routine and wait for her gaze. In case Miss Jane is out or has fallen asleep, it can anticipate where it should be when she returns or wakes up. In addition, the lights within the room are keyed off of the same input data and adjust automatically as she moves about her day so that she does not have to work the switches in the rooms. One wall in the main living space performs a keyword search of the digital TV Guide and understands what types of shows Miss Jane likes. When it identifies shows in certain genres that she likes, it changes color to passively suggest to her that something is on that she may like to watch. The effect of the wall would be similar to the Detroit Airport's Colorful Light Tunnel.

This scenario includes several features, each of which would be appropriate as the scope of Project 3. We are most interested in the learning problem and are interested in using Project 3 to explore neural networks. We would like to work on the algorithm that will allow a television to track position and learn a routine. The other part that interests us is focusing on how the television distinguishes when the user is gazing at it as a way to prime it for use. Having in-house experience with either of these issues will have broader applicability and future usefulness as we move forward with our endeavors in the IMSA lab.

Again, your thoughts are greatly appreciated.

Apoorva and Joe