I've allready written this on the pd-mailinglist. I thought maybe you forum-guys could help, too...

Hello everyone,

I'm currently working on a project called motion design
in which we want to motion track a person walking
trough a prepared room, than taking this data and
generate a spline out of it.
Our aim is that we can create a "virtual instrument" out of the motiontracking data.

Do you think this is possible with pd?
Any Ideas how to do it?

We have 6 Months to get this thing right, so we have plenty of time to learn pd and other related stuff.

Thanks in advance
Martin

Well, how are you going to be tracking the motion? Video, pressure sensors, lasers? Are the 'users' aware of their contribution?

PD can certainly make use of all kinds of interesting data. Preparing and interpolating from raw data files can be done. The most difficult part is translating 4d (x,y,z,t) spatial info into sound, which is usually conceived as only possessing two dimensions(amplitude, time).

Hey, I'm workin on that project as well.

We thougt of tracking via video only and want to use 2 Cameras. From that data we need to figure out the person's 3D coordinates in the prepared room.
Do you believe pd can handle that?
I've found some patches which track motion in 2D (gridflow) but our project seems far more complex.

Once we get the position and can follow the probant's motion, the sound-calculating shouldn't be too complicated.

best regards
from Bremen

As beetleskin said we'll use 2 cameras to track the motion. An other project hast worked on tracking the motion this way on which we want to build. The "users" are aware of what's going on. Do you know any projects in which "3d-tracking" has been done?

The part where we want to get sound out of it is not so hard. We just want the tracked motion to create a hollow body which will resonate when someone is "hitting" it.

probably OpenCV has something for that, there's some partial support for OpenCV for all 3 video externals (gridflow, Gem via pix-opencv, pdp via pdp-opencv) but I don't know if any of them have exactly what you need

You could:

Give each person an infrared 'throwie' that would stick to them, or even have them wear white tshits.

Analyze the video with pix_data. You'll need to decide on your desired resolution, and build a iterative XY 'grid counter ' which outputs XY position from 0...1, read the help file if that sounds strange.

Find the 'whitest' grid point per frame, by comparing the output of pix_data's right outlet, and save it to a table.
(this is 2d, repeat for second camera.)
...

Just and Idea, if you want me to hack together an example, let me know.