PURPOSE: To execute attribute judgement with high reliability by obtaining the total value of the output value of a membership function concerning all the feature values with respect to all the standard patterns and discriminating the the standard pattern with a highest total value so as to execute attribute judgement.
CONSTITUTION: Mesh pattern data (f) which should be judged is inputted to a feature value extraction circuit 3 so as to extract a necessary feature value (j). The feature value (j) is inputted to a membership function comparing circuit 4 to output a contributing value (l). This is similarly executed with respect to prescribed seven feature values. Next, each distributing value (l) is inputted to a distributing value summing circuit 5 to output a total distributing value (m). In this case, through the use of a fuzzy inference method, a distributing value to the membership function of each character is obtained with regard to the seven feature values, the total distributing value is obtained along an abscissa by each character and a character with a largest total distributing value is set to be the judgement result at last. Setting feature data properly like this improves the rate of the correct answers of attribute judgement so as to execute attribute judgement with high reliability.
ISHII SHOICHI
TEJIMA MITSUTAKA
