Pattern recognition apparatus, method thereof, and program product therefor

摘要

When a feature vector is converted to a reduced vector, a converting unit samples N components of interest from the M components of the feature vector, executes the process of calculating one component of the reduced vector from the N components of interest by d times to create the d-dimensional reduced vector and, the converting unit (1) excludes the components within a predetermined distance D in the same row as the previous component of interest sampled at the previous sampling, (2) excludes the components in the same column as the previous component of interest including the component k rows apart and within the distance (D−k) from the component k rows apart, and (3) samples the component of interest of this time from the remaining components after exclusion when sampling the component of interest.

主权项

1. A pattern recognition apparatus comprising:
an acquiring unit configured to acquire an image as a pattern; an extracting unit configured to perform a raster scan for W columns in the lateral direction and K rows in the vertical direction on the pattern and extract an M-dimensional feature vector having M components; a converting unit configured to reduce the dimensions of the M-dimensional feature vector to convert into a d-dimensional (M>d>0) reduced vector; a memory to store a set of reduction learned vectors learned in advance so as to correspond to each of categories; a recognizing unit configured to calculate a degree of similarity between the converted reduced vector and the reduction learned vector for each of the categories; and a determining unit configured to determine the category corresponding to the reduction learned vector similar to the reduced vector to be the category which the pattern belongs to on the basis of the degree of similarity, wherein the converting unit samples N (M>N>0) components of interest from the M components of the feature vector, executes the process of calculating one component of the reduced vector from the N components of interest by d times to create the d-dimensional reduced vector, and the converting unit (1) excludes the components that are arranged in same row as previously-designated component of interest, which has been sampled at previous sampling, and are within a predetermined distance D from the previously-designated component of interest, (2) excludes the components that are arranged in same column as the previously-designated component of interest, and are within a predetermined distance D from the previously-designated component of interest, (3) excludes any of the components, which is distanced from any of currently excluded component in the same row, by k rows (K>k>0), and is within a distance D−k from any of currently excluded component in the same column, and (4) samples currently-designated component of interest from remaining components that are remained after accumulated exclusion of components from first exclusion through current exclusion.