Proper recognition of a position can be achieved even when a predetermined part of a player moves rapidly. An image of a player is captured by a camera unit, and from this image the head position of the player is recognized, so that the game proceeds based on the recognized result. At this time, a base position group is selected in principle from a difference between the latest and immediately preceding player images within a selection range corresponding to the previously recognized position (S207), and the selection range is changed exceptionally to reselect the base position group from the difference (S210).