JP2957097B

PURPOSE: To perform a discrimination analysis with higher precision and to improve the recognition precision when an unknown pattern is recognized by using a dictionary obtained by repeating learning, by increasing the number of rival patterns at the time of the learning of the dictionary. CONSTITUTION: A discrimination analysis part 6 performs a discrimination- analyzing operation by using feature vector of a rival pattern corresponding to a noticed pattern and feature vector of the noticed pattern held in a noticed pattern holding part 7. Parameters for discriminating the obtained rival pattern and noticed pattern are stored in a discrimination dictionary 8. A combination parameter determination part 9 determines optimum combination parameters for combining a recognition function and a discrimination function that are prescribed by a identification dictionary 3 and a discrimination dictionary 8. In a learning dictionary 10, parameters regarding the learning distance function when the identification function and discrimination function are combined with each other on the basis of the combination parameters determined by the combination parameter determination part 9 are written.