I mean, if they found a way with similar data, there should be a way.
Kinect sees both but can’t distinguish which is which.
Say, if an arm swing inward (angular acceleration) towards the chest faster than the other, it should be the inner one right?
Another scenario would be dual yielding guns perpendicular to sensor. You know which one gets occlude first right? that would be the furthest one from the sensor.
Your skeleton has arm/leg lengths. You can calculate where it is approx. when the sensor doesn’t see it right?
I might be simplifying the problem too much. But i’m just throwing ideas around.