In this paper, a method is proposed to discriminate simple geometric shape patterns such as a triangle, a square, a regular pentagon, a regular hexagon, a circle, a rhombus, a trapezoid etc.. The pattern-classification value is used as a new parameter of feature extraction of shape pattern. The pattern-classification value is detected by projecting the shape pattern perpendicularly on the reference line. In the projected pattern of the shape, let A_x stand for the largest width on the reference line, A_y stand for the largest height from the reference line and A_a stand for the area of the pattern. The pattern-classification value is defined as the ratio of A_a to the product of A_x and A_y.
Some simple shapes are theoretically analyzed. The general characteristics of the pattern-classification value are that it is constant with a circle and regular polygons except multiple-of-four polygons. However, in the multiple-of-four polygons, the pattern-classification value changes by changing the postural angle and maintains a 180° period of cycle. Therefore, these characteristics can be utilized in the pattern discrimination of simple shapes. Some of the experimental results show a good a-greement with those of the theoretical analysis.