Towards optimization of mid-air gestures for in-vehicle interactions