Real-Time Multi-Sensor Hand Tracking System: Achieving <33ms Latency on $1000 Budget

I want to build a real-time system that would take input from a sensor and plot it's position on the screen. I would have many sensors and I would have to integrate all their incoming signals and would have to decode what their position on the screen is. Then I would map the points to something like a hand or a stick. The need for this system to be running as fast as possible should be emphasized. For the person to not be able to notice any difference between the motion of his/her fingers and the image on the screen. Is this possible? What kind of wire connections would I be looking at to maximize the speed that the signal travels at. Can the display update fast enough? Since human vision functions at around 30Hz, this should all happen in 0.33 seconds or 33ms. Is that kind of speed achievable with funds of about $1000? Lots of questions, but it's something I've been wondering about for a while, Thanks!

Hey Tim,

I don't have too many ideas on how you can or would like to approach this, but it sounds like your biggest concern is cheap yet powerful computing. Have you looked into the raspberry pi? It seems like that would be cheap but powerful and fit your needs well.

In terms of wiring and displays, I think you should be absolutely fine, as long as your backend processing can keep up. That's where your real chokepoint will be.

Good luck!
Josh

Note: 30 Hz would be every .033 seconds, an order of magnitude faster. Also, though the requirements are also little vague, have you considered a touch screen input?

The way I had it designed (in my mind lol) Is by having an PIC chip for every 2 sensors. I didn't really think that CPU will be my bottleneck. I don't have much experience with this but I though all I had to do was sort of "map" the point giving to me by the sensor into a 3d coord system, then apply a perspective and view transform and pass it to a rasterizer. That would be about 3 to 4 matrix opperations. I had expected the display and the rasterizer to take up most of the time. Can you explain this some more?

No worries, I understand the mental design process. I guess a lot depends on your input for the 3D coordinate system. If you're looking at a video input, then you will definitely need more power. But I think that either way the Pi is a good idea. It gives a good foundation so that you don't need to figure out the display from the bottom up. Unfortunately, I don't have any specific help on the issue. Hopefully someone else will pipe up with a little more experience. This should be interesting, though. Please put it on the projects here, I'd like to see how this progresses.