A detection area is set in a three-dimensional space in which a subject exists. When an actual hand enters the detection area, coordinate points (white and black dots) represented by pixels making up a silhouette of the hand in a depth image enter the detection area. In the detection area, a reference vector is set that shows the direction which the hand should face relative to the shoulder as a reference point. Then, an inner product between two vectors, a vector from the reference point to each of coordinate points and the reference vector, is calculated, followed by comparison between the inner products. Positions of coordinate points whose inner products are ranked high are acquired as the position of tips of the hand.