hey !

I've put my hands on a kinect, and thanks to Matthias Kronlachner I've pluged it to PureData with pix_openni.

So I receive points with XYZ translations,

and I want to transform that in rotations for a rig.

I'm no expert with such things. but I tryed something out with trigonometry :

a being the hand and b being the shoulder

BC = (a.x-b.x);
AB = (a.y-b.y);
AC = sqrt((AB*AB)+(BC*BC));
BD = AC - AB;
YY = atan(AB/BC)+atan(BD/BC)
X = (YY*180/PI)*2;

this gives me in degree from 0 to 360 the angle of my arm, I then just give him XY, XZ and ZY translation and get RX RY RZ in return.

But when I apply it as a rotation to a cube, movements don't match ...

So I guess that's not the right technique. I'm reading something about matrices, but I don't understand it quite well ...

so maybe somebody can tell me if I'm on the right way, and maybne explain me a bit

thanks for reading,
cheers

actually I just found that

http://en.wikipedia.org/wiki/Gimbal_lock

wich is I think my problem, I'm trying now to understand how to write a solution in PD

http://pastebin.com/RQgwaC1J

I made a patch to fix gimbal lock on PD but I still have wrong moves,

can someone telle me if the patch is wrong ?

or if after gimbal fix following rotations

X 160 Y 90 Z 267

with Y to be the base

you get

X 0 Y 90 Z 106

thanks

well I found a euler/quaternion/euler calculator online and after compare his results with mine, it appears that the part I have wrong is the RZ in case of gimbal lock

maybe someone can telle me wich part is wrong

RX = ((sin_X*sin_Y)*sin_Z)+(cos_X*cos_Z);
RY = -(sin_X*sin_Y)*cos_Z+cos_X*sin_Z;

Z = Math.atan2(RY,RX);
Z *= RADIANS;

cheers