Did some experiment to improve our VR prototype today.
According to feedbacks, the body was really annoying, so I want to find a way to improve it.
I tried to use could-points data from depth data, but it didn't work well because once player not in the Kinect scan area, the image will lose some parts.
Then I tried to use pure joints points plus some particle effects, seems worked pretty well.
In this way, really cannot tell most jittering, because it always seems moving.... But it looks not bad in the game.
Also I post some depth data image(latter ones).