LEARNING SUPPORTING DEVICE AND PATTERN RECOGNITION DEVICE

摘要

PROBLEM TO BE SOLVED: To realize the improvement of a recognizing time, a learning time and recognizing performance and to prevent the deterioration of recognizing performance caused by the change of environment by realizing the automatic selection of learning data and the selection of an optimum feature value in each optimum recognition category in pattern recognition using a neural network. SOLUTION: A feature value arithmetic part 102 calculates the feature value with respect to each image data 101. A static analytic part 105 executes cluster analysis with respect to each set of feature values corresponding to each image data to sort the set of the feature value of each into plural clusters corresponding to each category of the image data 101 to select the set representing each sorted cluster as learning data. The set of these feature values is normalized by a normalizing part 103 and then used for learning at a neural network part 104.