Pattern recognition method, pattern check method and pattern recognition apparatus as well as pattern check apparatus using the same methods
First Claim
Patent Images
1. A method of pattern recognition, said method comprising the steps of:
- a first step for finding feature-extraction-matrix AF of pattern set A1 and feature-extraction-matrix BF of pattern set B1, said matrices AF, BF maximizing between-class scatter and minimizing within-class scatter with respect to training pattern sets A1, B1 respectively, which are sets of pair of patterns obtained from a same object but on different conditions;
a second step for calculating feature-amount fB1i using the feature-extracting-matrix BF with respect to typical pattern B1i of respective classes “
i”
of the training pattern set B1;
a third step for retaining a set {fB1i} of the feature amount fB1i and the matrices AF, BF in a referential database; and
a fourth step for determining an element, among the feature amount fB1i retained in the referential database, most similar to a feature amount fA2j extracted from the matrix AF applied to input pattern A2j.
1 Assignment
0 Petitions
Accused Products
Abstract
With respect to two pattern sets obtained on different conditions, a feature-extraction matrix, which maximizes between-class scatter and minimizes within-class scatter, is found respectively. A first feature amount is calculated using one of the feature-extraction matrices. The first feature amount and the two matrices are retained in a referential database. A pattern is determined to a second feature amount—extracted by applying another feature-extraction matrix to a pattern input—by extracting a most similar element out of the first feature amount retained in the referential database.
12 Citations
26 Claims
-
1. A method of pattern recognition, said method comprising the steps of:
-
a first step for finding feature-extraction-matrix AF of pattern set A1 and feature-extraction-matrix BF of pattern set B1, said matrices AF, BF maximizing between-class scatter and minimizing within-class scatter with respect to training pattern sets A1, B1 respectively, which are sets of pair of patterns obtained from a same object but on different conditions;
a second step for calculating feature-amount fB1i using the feature-extracting-matrix BF with respect to typical pattern B1i of respective classes “
i”
of the training pattern set B1;
a third step for retaining a set {fB1i} of the feature amount fB1i and the matrices AF, BF in a referential database; and
a fourth step for determining an element, among the feature amount fB1i retained in the referential database, most similar to a feature amount fA2j extracted from the matrix AF applied to input pattern A2j. - View Dependent Claims (2, 3, 4, 5, 6)
-
-
7. A method of pattern check, said method comprising the steps of:
-
a first step for finding feature-extraction-matrix AF of pattern set A1 and feature-extraction-matrix BF of pattern set B1, said matrices AF, BF maximizing between-class scatter and minimizing within-class scatter with respect to the training pattern sets A1, B1 which are sets of pair of patterns obtained from a same object but on different conditions;
a second step for calculating feature amounts “
fa”
, “
fb”
of input patterns “
a”
, “
b”
using the matrices AF, BF; and
a third step for determining whether or not the input patterns “
a”
, “
b”
derive from a same class based on similarity measure of the feature amounts “
fa”
, “
fb”
. - View Dependent Claims (8, 9, 10, 11, 12)
-
-
13. A pattern recognition apparatus comprising:
-
(a) pattern input means for obtaining training-pattern-sets A1, B1, which are sets of pair of patterns, from a same object but on different conditions;
(b) a feature-extraction-matrix generating unit for finding feature-extraction-matrix AF of the pattern set A1 and feature-extraction-matrix BF of the pattern set B1, both matrices AF, BF maximizing between-class scatter and minimizing within-class scatter with respect to the pattern sets A1, B1;
(c) a feature-amount calculating unit for calculating a feature amount fB1i using the feature-extraction-matrix BF with respect to typical pattern B1i of each class “
i”
of the training pattern set B1;
(d) transfer means for transferring a set {fB1i} of the feature amount fB1i and the feature-extraction-matrix BF to referential database FB1; and
(e) optimum matching determination means for selecting an element most similar to feature amount fA2j from the database by applying the feature-extraction-matrix AF to pattern A2j input, the feature amount fA2j being extracted after applying the matrix AF to the pattern A2j. - View Dependent Claims (14, 15, 16, 17)
-
-
18. A pattern check apparatus comprising:
-
(a) a pattern inputting unit for obtaining training pattern sets A1, B1, which are sets of pair of patterns, from a same object but on different conditions;
(b) a feature-extraction-matrix generating unit for finding feature-extraction-matrix AF of the pattern set A1 and feature-extraction-matrix BF of the pattern set B1, both matrices AF, BF maximizing between-class scatter and minimizing within-class scatter with respect to the pattern sets A1, B1;
(c) a feature-extracting-unit for calculating respective feature amounts “
fa”
, “
fb”
using the feature-extracting-matrices AF, BF from input pattern “
a”
, “
b”
supplied from said pattern input unit;
(d) an identity determination unit for determine whether or not the input patterns “
a”
, “
b”
are derived from the same class based on similarity measure of the feature amounts “
fa”
, “
fb”
. - View Dependent Claims (19, 20, 21, 22)
-
-
23. A recording medium which records a program of recognizing a pattern by a computer, said program carrying out the processes of:
-
(a) a process for finding feature-extraction-matrix AF of pattern set A1 and feature-extraction-matrix BF of pattern set B1, said matrices AF, BF maximizing between-class scatter and minimizing within-class scatter with respect to the training pattern sets A1, B1 which are sets of pair of patterns obtained from a same object but on different conditions;
(b) a process for calculating feature amount “
fB1i”
using the feature-extracting-matrix BF with respect to typical pattern “
B1i”
of respective classes “
i”
of the training pattern set B1;
(c) a process for retaining a set {fB1i} of the feature amount fB1i and the matrices AF, BF in a referential database; and
(d) a process for determining an element, among the feature amount fB1i retained in the referential database, most similar to a feature amount fA2j extracted from the matrix AF applied to input pattern A2j. - View Dependent Claims (24)
-
-
25. A recording medium which records a program of checking a pattern by a computer, said program carrying out the processes of:
-
(a) a process for finding feature-extraction-matrix AF of pattern set A1 and feature-extraction-matrix BF of pattern set B1, the matrices AF, BF maximizing between-class scatter and minimizing within-class scatter with respect to the training pattern sets A1, B1 which are sets of pair of patterns obtained from a same object but on different conditions;
(b) a process for calculating feature amounts “
fa”
, “
fb”
of input patterns “
a”
, “
b”
using the matrices AF, BF; and
(c) a process for determining whether or not the input patterns “
a”
, “
b”
derive from a same class based on similarity measure of the feature amounts “
fa”
, “
fb”
. - View Dependent Claims (26)
-
Specification
-
- Resources
Litigation Campaign Assessment
Thank you for your request. You will receive a custom alert email when the Litigation Campaign Assessment is available.
×
-
Current AssigneeMatsushita Electric Industrial Company Limited (Panasonic Holdings Corporation)
-
Original AssigneeMatsushita Electric Industrial Company Limited (Panasonic Holdings Corporation)
-
InventorsNagao, Kenji
-
Primary Examiner(s)MARIAM, DANIEL G
-
Application NumberUS09/874,199Publication NumberTime in Patent Office1,372 DaysField of Search382/115, 382/118, 382/159, 382/170, 382/181, 382/190, 382/192, 382/195, 382/197, 382/209, 382/218, 382/224, 382/225, 382/228US Class Current382/209CPC Class CodesG06F 18/213 Feature extraction, e.g. by...
