METHOD FOR FEATURE SELECTION FOR PATTERN RECOGNITION AND APPARATUS THEROF

摘要

PURPOSE: A feature data selection method and a feature data selection apparatus thereof are provided to remove un-uniformity between class-feature information and feature-feature information. CONSTITUTION: A data normalization part(11) normalizes feature data which is extracted from inputted data. A data converting part(13) converts normalized successive type feature data into discrete type feature data. A first feature data selection part(15) calculates the correlation of the normalized discrete type feature data. The first feature data selection part selects first feature data. A second feature data selection part(17) calculates the potential of the feature data which is not selected.