RECOGNITION APPARATUS, REAL NUMBER MATRIX DECOMPOSITION METHOD, AND RECOGNITION METHOD

摘要

PROBLEM TO BE SOLVED: To provide a recognition apparatus capable of quickly recognizing a target that entails the inner product calculation between a binarized feature vector and each of a plurality of real number vectors.SOLUTION: An object recognition apparatus 30 comprises: a binary code conversion unit 33 that acquires a binarized feature vector x; a parameter-matrix decomposition unit 35 decomposing a real number matrix configured from a plurality of real number vectors qto qwhich serves as a dictionary of a plurality of classes for identifying the feature vector x into a sum of a plurality of MC(where i=1 to N and N is a natural number not smaller than 2); and a linear SVM identification unit 36 that calculates a product xMbetween the feature vector x and a basic matrix Mas calculation of an inner product between the feature vector x and each of the real number vectors qto q, and furthermore calculates the product between the product xMand a corresponding coefficient matrix C, thereby identifying each class for the feature vector x.SELECTED DRAWING: Figure 12