that or extracted normal maps that blend in and out (or even animated maps) tho that is like the next level. First you want a facial system ingame either with just blend targets or bones.
Just yesterday I was able to add my full bone based facial rig into the engine. Im currently applying it via a separate facial state machine and using blendAdditive in the animGraph to merge facial and body animations.
By having a bone based rig and regular skeletal animation clips for facial animation its then easy to drive off blendTargets and facial wrinkle maps in those clips with notifies or some system im yet not aware of. Also you can then create facial animations that is really targeted to the speech file in your dccs (maya,max,blender) with more than just blunt lipsync and have them nice and synchronized on execution ingame.
So how are you planning on making your face rig? bones only? A combination of bones for broad areas and blend targets for details? Or completely in blend targets?