Apparatus and method for robust pattern recognition
First Claim
1. A method for robust pattern recognition, comprising the steps of:
- (a) generating N sets of feature vectors x1, x2, . . . xN from a set of observation vectors which are indicative of a pattern which it is desired to recognize, at least one of said sets of feature vectors being different than at least one other of said sets of feature vectors and being preselected for purposes of containing at least some complimentary information with regard to said at least one other of said sets of feature vectors; and
(b) combining said N sets of feature vectors in a manner to obtain an optimized set of feature vectors which best represents said pattern, said combining being performed in accordance with the equation;
p(x1, x2, . . . xN|sj)=f—
n{K+[w1p(x1|sj)q+w2p(x2|sj)q+. . . +wNp(xN|sj)q]1/q}where;
f—
n is one of an exponential function exp( ) and a logarithmic function log( ), sj is a label for a class j, N is greater than or equal to 2, p(x1, x2, . . . xN|sj) is conditional probability of feature vectors x1, x2, . . . xN given that they are generated by said class j, K is a normalization constant, w1, w2, . . . wN are weights assigned to x1, x2, . . . xN respectively according to confidence levels therein; and
q is a real number corresponding to a desired combination function.
3 Assignments
0 Petitions
Accused Products
Abstract
N sets of feature vectors are generated from a set of observation vectors which are indicative of a pattern which it is desired to recognize. At least one of the sets of feature vectors is different than at least one other of the sets of feature vectors, and is preselected for purposes of containing at least some complimentary information with regard to the at least one other set of feature vectors. The N sets of feature vectors are combined in a manner to obtain an optimized set of feature vectors which best represents the pattern. The combination is performed via one of a weighted likelihood combination scheme and a rank-based state-selection scheme; preferably, it is done in accordance with an equation set forth herein. In one aspect, a weighted likelihood combination can be employed, while in another aspect, rank-based state selection can be employed. An apparatus suitable for performing the method is described, and implementation in a computer program product is also contemplated. The invention is applicable to any type of pattern recognition problem where robustness is important, such as, for example, recognition of speech, handwriting or optical characters under challenging conditions.
42 Citations
44 Claims
-
1. A method for robust pattern recognition, comprising the steps of:
-
(a) generating N sets of feature vectors x1, x2, . . . xN from a set of observation vectors which are indicative of a pattern which it is desired to recognize, at least one of said sets of feature vectors being different than at least one other of said sets of feature vectors and being preselected for purposes of containing at least some complimentary information with regard to said at least one other of said sets of feature vectors; and
(b) combining said N sets of feature vectors in a manner to obtain an optimized set of feature vectors which best represents said pattern, said combining being performed in accordance with the equation;
p(x1, x2, . . . xN|sj)=f—
n{K+[w1p(x1|sj)q+w2p(x2|sj)q+. . . +wNp(xN|sj)q]1/q}where;
f—
n is one of an exponential function exp( ) and a logarithmic function log( ),sj is a label for a class j, N is greater than or equal to 2, p(x1, x2, . . . xN|sj) is conditional probability of feature vectors x1, x2, . . . xN given that they are generated by said class j, K is a normalization constant, w1, w2, . . . wN are weights assigned to x1, x2, . . . xN respectively according to confidence levels therein; and
q is a real number corresponding to a desired combination function. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20)
-
-
21. An apparatus for robust pattern recognition, said apparatus comprising:
-
(a) a feature vector generator which generates N sets of feature vectors x1, x2, . . . xN from a set of observation vectors which are indicative of a pattern which it is desired to recognize, at least one of said sets of feature vectors being different than at least one other of said sets of feature vectors and being preselected for purposes of containing at least some complimentary information with regard to said at least one other of said sets of feature vectors; and
(b) a feature vector combiner which combines said N sets of feature vectors in a manner to obtain an optimized set of feature vectors which best represents said pattern, said combining being performed in accordance with the equation;
p(x1, x2, . . . xN|sj)=f—
n{K+[w1p(x1|sj)q+w2p(x2|sj)q+. . . +wNp(xN|sj)q]1/q}where;
fn is one of an exponential function exp( ) and a logarithmic function log( ), sj is a label for a class j, N is greater than or equal to 2, p(x1, x2, . . . xN|sj) is conditional probability of feature vectors x1, x2, . . . xN given that they are generated by said class j, K is a normalization constant, w1, w2, . . . wN are weights assigned to x1, x2, . . . xN respectively according to confidence levels therein; and
q is a real number corresponding to a desired combination function. - View Dependent Claims (22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40)
-
-
41. A program storage device readable by machine, tangibly embodying a program of instructions executable by the machine to perform method steps for robust pattern recognition, said method steps comprising:
-
(a) generating N sets of feature vectors x1, x2, . . . xN from a set of observation vectors which are indicative of a pattern which it is desired to recognize, at least one of said sets of feature vectors being different than at least one other of said sets of feature vectors and being preselected for purposes of containing at least some complimentary information with regard to said at least one other of said sets of feature vectors; and
(b) combining said N sets of feature vectors in a manner to obtain an optimized set of feature vectors which best represents said pattern, said combining being performed in accordance with the equation;
p(x1, x2, . . . xN|sj)=f—
n{K+[w1p(x1|sj)q+w2p(x2|sj)q+. . . +wNp(xN|sj)q]1/q}where;
fn is one of an exponential function exp( ) and a logarithmic function log( ), sj is a label for a class j, N is greater than or equal to 2, p(x1, x2, . . . xN|sj) is conditional probability of feature vectors x1, x2, . . . xN given that they are generated by said class j, K is a normalization constant, w1, w2, . . . wN are weights assigned to x1, x2, . . . xN respectively according to confidence levels therein; and
q is a real number corresponding to a desired combination function.
-
-
42. A method for robust pattern recognition, comprising the steps of:
-
(a) generating N sets of feature vectors x1, x2, . . . xN from a set of observation vectors which are indicative of a pattern which it is desired to recognize, at least one of said sets of feature vectors being different than at least one other of said sets of feature vectors and being preselected for purposes of containing at least some complimentary information with regard to said at least one other of said sets of feature vectors; and
(b) combining said N sets of feature vectors in a manner to obtain an optimized set of feature vectors which best represents said pattern, said combining being performed via one of;
a weighted likelihood combination scheme wherein a set of weights are assigned to corresponding likelihoods from each of said N sets of feature vectors; and
a rank-based state-selection scheme wherein that one of said N sets of feature vectors for which a corresponding one of said likelihoods has a highest rank is selected.
-
-
43. An apparatus for robust pattern recognition, said apparatus comprising:
-
(a) a feature vector generator which generates N sets of feature vectors x1, x2, . . . xN from a set of observation vectors which are indicative of a pattern which it is desired to recognize, at least one of said sets of feature vectors being different than at least one other of said sets of feature vectors and being preselected for purposes of containing at least some complimentary information with regard to said at least one other of said sets of feature vectors; and
(b) a feature vector combiner which combines said N sets of feature vectors in a manner to obtain an optimized set of feature vectors which best represents said pattern, said combining being performed via one of;
a weighted likelihood combination scheme wherein a set of weights are assigned to corresponding likelihoods from each of said N sets of feature vectors; and
a rank-based state-selection scheme wherein that one of said N sets of feature vectors for which a corresponding one of said likelihoods has a highest rank is selected.
-
-
44. A program storage device readable by machine, tangibly embodying a program of instructions executable by the machine to perform method steps for robust pattern recognition, said method steps comprising:
-
(a) generating N sets of feature vectors x1, x2, . . . xN from a set of observation vectors which are indicative of a pattern which it is desired to recognize, at least one of said sets of feature vectors being different than at least one other of said sets of feature vectors and being preselected for purposes of containing at least some complimentary information with regard to said at least one other of said sets of feature vectors; and
(b) combining said N sets of feature vectors in a manner to obtain an optimized set of feature vectors which best represents said pattern, said combining being performed via one of;
a weighted likelihood combination scheme wherein a set of weights are assigned to corresponding likelihoods from each of said N sets of feature vectors; and
a rank-based state-selection scheme wherein that one of said N sets of feature vectors for which a corresponding one of said likelihoods has a highest rank is selected.
-
Specification