If you want to manipulate your model in three dimensions, you will need to break up your model into "planes of movement" and use several bone layers containing one or more objects. If you rotate and translate the bone layers in 3D, you can simulate 3D movement of your model. If you put the center of a bone layer at the hinge point of a limb, you should have a very flexible setup, because you can swing the arms and legs (and head) in 3D very easily using layer rotation of the bone layers.
Here is a roughly animated example I made on YouTube:
http://www.youtube.com/watch?v=EUuBF5mxVuU
I used 2D vector layers, but you could just as well have used 3D objects or image layers. I didn't fine tune the animation with switch layers, to switch to the appropriate vector layer for a particular viewing angle. If I had used this refinement for both the collar and the left arm the result would have been more convincing.
Here's the setup in layers and a screenshot showing the setup from another angle. Notice in the screenshot, the legs are in a different plane, rotated by 45 and -45 degrees, respectively, and the head is translated in the positive z direction (further to the front than the torso). I have used layer binding to bind the layers Head, Torso, L leg, R leg, L arm and R arm and in the bone layers L Leg, R leg, L arm and R arm I used normal region binding. In the bone layer Someone, I used Depth Sorting by depth and true distance.


I realize this is not a very sophisticated setup and many variations are possible. I'll leave that up to you to discover.