I'm sure many people have thought about the behavior of first person mouse-look before, so how about a discussion? I had a lot of thoughts about advanced mouse-look models during last summer (2012). I thought I'd share my ideas. Please note that these ideas may be extended to any control schemes of first person viewing models (so, not limited to mouse-look control).
When I roll my eyeballs around, the motion I notice really doesn't seem anything at all like a spherical rotation being applied to a view tranformation. The motion seems more like my peripheral view is shifting. So it's like the projection matrix is merely offset. You can take D3DXMatrixPerspectiveOffCenterLH
's functionality for concrete instance (just for a common reference, in case you really don't understand what I mean).
Obviously, any form of mouse-look interpretation which distinguishes movement between the eyeballs, head and the character person's entire torso/body will have a greater danger of being harmful to the gameplay experience of fast paced action games such as first-person shooters. I've been thinking on various ways by which mouse-look control may be interpeted most accurately by 3 domains of user-intention. If the player is walking or running, whole-torso movement twisting control will be most influenced by mouse-look activity (when running; this weight will be very high, 95 - 100%). Only when idle, if the player makes subtle/trivial movements with their mouse -- let's say within an 8 pixel radius in screen space -- the eyeball movement power may be concentrated even somewhete from 40 to 50% weight. Between the extremes of torso-twisting and acute eye movement, head movements can account for remaining behavior (though the weighting will be much nearer to eyeball movement than to the effect of torso twisting).
Also, the particular ways by which orientational components are affected through mouse-look can vary. Think of Euler angles. It's not very easy to 'pitch' or 'roll' your torso, though you may bend your back a little. The yaw of a person's torso usually will be a vector that is often nearly (or exactly) tangent to the ground which this entire person is standing upon. For heads, 'yaw' probably will need to be the most limited factor, though 'pitch' can have the most freedom while 'roll' is to have slightly less. For eyeballs, I'm guessing that 'roll' may only slightly get sheared when looking towards peripheral corners, but both 'yaw' and 'pitch' will probably be provided roughly the same degree of freedom.
Yet all of this mouse-look behavior is only determined on behalf of screen space-based mouse activity. What about a 'world space' interpretation? If the mouse movement vector -- a vector that is relatively designated as a screen-space offset located in the center of perspective which measures the horizontal and vertical deltas of mouse movement -- is ray-casted into the game scene, then what kind of influences should its sampling cause to the interpretation of mouse-look control?
I may easily write and speculate more on this, but I'm stopping here to get this topic posted and go to sleep.
So hearken, this community of might and glory; you geniuses out there! Please add your thoughts!