Culture: Movie: Making
RFID motion capture   (+2, -2)  [vote for, against]
USing RFID chips placed in a leotard outfit for motion capture

This is an idea for a poor-man's motion capture set up.

You have the actors have full body leotards, thight-fighting outfits. In certain places, you place RFID chips, like in the wrist, elbow, shoulder -- each point where you want to track location. Then, put three or more receivers in a ring around the performance area to triangulate the position of the RFID chips.

Of course, you would need a lot of tracking points, so in order to solve this, you have a each chip on the same frequency. You have the program deduce the location by figuring out the centers of interference of the radio waves, like in the image provided.
-- lawpoop, Dec 14 2008

Wave interference http://upload.wikim...en/9/94/Interf2.png
The visual guide to figuring out where the chips are located [lawpoop, Dec 14 2008]

More interference http://www.asci.org...ve_Interference.jpg
yet more interference [lawpoop, Dec 14 2008]

Triangulation http://www.coyleweb.net/triangulation.jpg
A visual description of triangulation. Each dot is a circle represents a receiver. The intersection of the circle would be the source of the signal; in this case, the RFID chip. [lawpoop, Dec 16 2008]

Current motion capture system http://www.animatio....org/pics/mocap.jpg
This is the system that current motion capture programs use. [lawpoop, Dec 18 2008]

I'm not convinced this would be possible. I'm sure RFID doesn't offer either direction or distance information. If you had a dense enough of receivers, you might get a low-res 2-d map.

Your interference example doesn't apply, I don't believe, because RFID doesn't operate off signal modulation. The receivers either read the chip or they don't.
-- phoenix, Dec 14 2008


Deducing the position based on electromagnetic interference is already how many existing systems in VR work (e.g., the Polhemus cube.)

(If you want something cheaper, the usual direction is from magnetic fields to optical - not the other way 'round!)
-- jutta, Dec 15 2008


[Phoenix] You are right, but you don't use the RFID information itself for either direction or distance.

First off, direction is not recorded. Animation is like a flip-book -- if you look at the cartoon cells of Mickey Mouse, you'll see that they are all still frames. There is no motion. When you show a series of still frames, the mind pastes together motion.

So, all you have to do is get a series of snapshots of points in space. The motion is provided by the actors. It's the same as if you were shooting a movie, except this is 3-D.

Second, the RFID doesn't give information about distance. But that's what the triangulation is for -- which is why you have 3 or more receivers. It's the same way they do cell phone triangulation. The cell phone itself cannot say where it is ( other than "here"). What happens is that 3 or more towers calculate the time it takes to get a signal back from a particular phone. When you have those three measurements, you can figure out where the item is located, by drawing a triangle. See link.

Finally, RFID chips do give a burst of radio information. We're not really interested in reading them as RFID chips; we just want to triangulate the source of the burst. The reason we use RFID is because they're cheap, and they're small enough to put on an actor's suit.
-- lawpoop, Dec 16 2008


One of the main problem I can see is the necessary speed of the signal processing.

To get centimetre precision of measurements, you'd need to measure at a resolution of at least 30 pico seconds. That's not impossible for a single RFID transmitter, but multiple RFID transmitters might make it more difficult.

And rather than using RFID, it might be easier to use a simple radio transmitter (with its own power source) that transmits a radio pulse at regular intervals.
-- xaviergisz, Dec 16 2008


Some years back there were some games made by Zowie Intertainment, which sensed the position of action figures on a playing surface. Each figure had a tuned LC circuit. I don't know what the limitations would be on doing something like that on a larger scale, but I thought the playsets worked pretty impressively.
-- supercat, Dec 17 2008


Okay, I'm not going to beat this horse but,
1) It's "phoenix". Oe (or œ if you like), not eo and,
2) "RFID doesn't give information about distance. But that's what the triangulation is for"
Triangulation is no good without direction information. RFID is omnidirectional, so the entire area of overlap between the receivers could contain the tag. That's what I meant by "low resolution". If you do some post-processing, you can work your way back to the probable origin of the signal, but I don't think you'd never really be sure - especially if you're trying to track many tags at once with just a few receivers. You'll end up with "blobs". They might move from frame to frame, but they'll still be blobs.

"It's the same way they do cell phone triangulation"
Cellular signals have timing information embedded in them. RFID doesn't.
-- phoenix, Dec 17 2008


[Phoenix] I apologize for spelling your name wrong. I corrected it.

I'd like to understand your criticism, if you can bear with me.

I think what I failed to mention is that the receivers are also transmitters. ( This is how RFID receivers work ) When a transmitter gives a burst, it can know at what time it gave the burst. The burst travels out to the RFID chip, energizes it, and then the chip emits a signal. The receiver receives the RFID signal, and it can know at what time it received the response. So it knows the total round trip time, from burst to response. Assuming that it took no time for the RFID to change the burst into a response, we can then conclude that the distance from the transmitter/receiver is 1/2 the time from burst to response. So, then you have a sphere of that distance around the receiver as to where the chip could be. If you have two receivers, then you can have two spheres, with an intersection. You would have a 3-D vesica pisces, so to speak. So you've whittled down the possible locations. Finally, you add a third receiver, and you should have a point where the three spheres intersect. If you want to get more accurate, you could add more receivers.

I think this makes sense. This is how Kepler calculated the transit of Venus, through triangulation. Venus didn't transmit timing information. How does direction figure into this? How does direction information turn a blob into a point?

I'm not talking about motion, I'm talking about where the chip is in a point at time. I don't care where the actor is moving the chip to, I only care where it is at 30 frames a second. Animation doesn't really move; it's just a flip-book, in essence.

So does that work for a single chip? Where did I go wrong?

If it works for a chip, then the problem is, we need to track several points to do motion capture of a human actor. You could use chips that transmit different frequencies, but would you run out of different frequencies to use? I don't know. But if so, my thought would be to use all the same frequency. In the image links that I posted, I can see the epicenters of the wave interference. I think you could write a program that could figure out the center of wave sources in those images.
-- lawpoop, Dec 18 2008


I think you might be confusing people. you say:

//...we can then conclude that the distance from the transmitter/receiver is 1/2 the time from burst to response//

but you also say:

//Really, this isn't that much different than current motion capture system, using reflective spheres. Instead of visible light, we're using radio waves.//

These are very different ways of finding the location of an object.

Also, I think you are confusing the idea by calling it triangulation. It is better to describe it as measuring the delay of a pulse at three receivers.

I think that your proposed system hasn't been used because accurately measuring the time difference between radio pulses (which travel at the speed of light) is hard.

To measure the position of a radio transmitter to centimetre accuracy would require a processor with a clock speed of 30GHz. This is approximately 10 times faster than the fastest commercially available processors.

It would be much easier to use transmitters that emit pulses of sound. Sound moves six orders of magnitude slower than light/radio, so can be dealt with by readily available signal processors.
-- xaviergisz, Dec 18 2008


"//...we can then conclude that the distance from the transmitter/receiver is 1/2 the time from burst to response//

but you also say:

//Really, this isn't that much different than current motion capture system, using reflective spheres. Instead of visible light, we're using radio waves.// "

OK, but to be fair, I only said that after Phoenix's criticism, so he couldn't have been confused by that. I've since removed it.
-- lawpoop, Dec 18 2008


"receivers are also transmitters"
You're right, and I knew that but wasn't thinking that way. What's the transmission rate on those things?

Hell, if you can make it work, more power to you!
-- phoenix, Dec 18 2008


[Phoenix] It's looking more and more possible! If I can just figure out the multiple motion points...
-- lawpoop, Dec 18 2008



random, halfbakery