GROUP DISCRIMINATION DEVICE AND GROUP DISCRIMINATION METHOD

摘要

PROBLEM TO BE SOLVED: To provide a new group discrimination device in which an SVM(support vector machine) learns a shape of a whole spectrum as well as a certain partial feature quantity.

SOLUTION: A group discrimination device 1 is provided with: a spectrum estimation means 2 for estimating a discrete power spectrum from time-space series data; a normalization means 3 for forming a normalized histogram by normalizing the spectrum; a histogram storage means 4 for storing the feature vector of the normalized histogram in association with the group; a discrimination hyperplane generation means 5 for classifying time-space series data for learning into a plurality of groups by using an SVM 6 for forming a discrimination hyperplane for classifying the feature vector in a feature space by using kernel functions; and a discrimination means 7 for discriminating a group to which the target time-space series data belongs by using the discrimination hyperplane.

COPYRIGHT: (C)2008,JPO&INPIT