| 摘要 |
PROBLEM TO BE SOLVED: To provide a discrimination method which enables more reduction in the calculation quantity of discrimination processing while keeping discriminating precision as compared with a conventional technology at the time of the construction of a discrimination rule using learning data, a discrimination device and a program. SOLUTION: The discrimination method employs a support vector machine method which uses a discrimination function which sets characteristic data for characterizing a discrimination target as input data to discriminate a kind to which the discrimination target belongs on the basis of the output value of the discrimination function. This discrimination method comprises a selection step of selecting a support vector which is used in the discrimination of the kind to which the discrimination target belongs on the basis of a distance on the characteristic quantity space of each vector and the discrimination target by predetermined processing from among a plurality of the support vectors constituting the discrimination function preliminarily calculated using the support vector machine method and a step of discriminating the discrimination target using the discrimination function constituted of the support vector selected in the selection step. COPYRIGHT: (C)2010,JPO&INPIT |
