Automatic recognition of a consistent message using multiple complimentary sources of information
First Claim
1. In a message recognition system, a method of transforming a consistent message into a message recognizable by the computer, the method comprising the steps of:
- (A) transforming the consistent message generated by a human in at least two formats into electrical signal representations of the consistent message;
(B) producing from said electrical signal representations of the consistent message a set of parameters for each said format;
(C) generating a likelihood score of recognition for each said set of parameters;
(D) using said sets of parameters to train a weighting coefficient for each of the at least two formats of the consistent message, wherein said step of training a weighting coefficient comprises the steps of(i) partitioning the consistent message in each said format into one or more subunits, wherein each subunit corresponds to a piece of the consistent message,(ii) grouping said subunits from each said format into a plurality of groups, wherein each group comprises one said subunit from each said format, and wherein each said subunit in one said group corresponds to the same piece of the consistent message,(iii) determining a likelihood score of recognition for each said group of subunits,(iv) determining a global score for the consistent message based on said likelihood scores of recognition, and(v) using said global score to determine said weighting coefficients,(E) generating a weighted expression based on said trained weighting coefficient and said likelihood scores of recognition; and
(F) selecting a candidate message unit that maximizes said weighted expression to transform said electrical signal representations of the consistent message into a computer recognizable message.
2 Assignments
0 Petitions
Accused Products
Abstract
A general approach is provided for the combined use of several sources of information in the automatic recognition of a consistent message. For each message unit (e.g., word) the total likelihood score is assumed to be the weighted sum of the likelihood scores resulting from the separate evaluation of each information source. Emphasis is placed on the estimation of weighing factors used in forming this total likelihood. This method can be applied, for example, to the decoding of a consistent message using both handwriting and speech recognition. The present invention includes three procedures which provide the optimal weighing coefficients.
351 Citations
28 Claims
-
1. In a message recognition system, a method of transforming a consistent message into a message recognizable by the computer, the method comprising the steps of:
-
(A) transforming the consistent message generated by a human in at least two formats into electrical signal representations of the consistent message; (B) producing from said electrical signal representations of the consistent message a set of parameters for each said format; (C) generating a likelihood score of recognition for each said set of parameters; (D) using said sets of parameters to train a weighting coefficient for each of the at least two formats of the consistent message, wherein said step of training a weighting coefficient comprises the steps of (i) partitioning the consistent message in each said format into one or more subunits, wherein each subunit corresponds to a piece of the consistent message, (ii) grouping said subunits from each said format into a plurality of groups, wherein each group comprises one said subunit from each said format, and wherein each said subunit in one said group corresponds to the same piece of the consistent message, (iii) determining a likelihood score of recognition for each said group of subunits, (iv) determining a global score for the consistent message based on said likelihood scores of recognition, and (v) using said global score to determine said weighting coefficients, (E) generating a weighted expression based on said trained weighting coefficient and said likelihood scores of recognition; and (F) selecting a candidate message unit that maximizes said weighted expression to transform said electrical signal representations of the consistent message into a computer recognizable message. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18)
-
-
19. A message recognition system for transforming a consistent message into a message recognizable by the computer comprising:
-
first transform means for transforming the consistent message generated by a human in at least two formats into electrical signal representations of the consistent message; production means for producing from said electrical signal representations of the consistent message a set of parameters for each said format; first generating means for generating a likelihood score of recognition for each said set of parameters; training means for using said sets of parameters to train a weighting coefficient for each of the at least two formats of the consistent message, wherein said training means comprises partitioning means for partitioning the consistent message in each said format into one or more subunits, wherein each subunit corresponds to a piece of the consistent message, grouping means for grouping said subunits from each said format into a plurality of groups, wherein each group comprises one said subunit from each said format, and wherein each said subunit in one said group corresponds to the same piece of the consistent message, first determining means for determining a likelihood score of recognition for each said group of subunits, second determining means for determining a global score for the consistent message based on said likelihood scores of recognition, and third determining means for using said global score to determine said weighting coefficients, second generating means for generating a weighted expression based on said trained weighting coefficient and said likelihood scores of recognition; and second transform means for selecting a candidate message unit that maximizes said weighted expression to transform said electrical signal representations of the consistent message into a computer recognizable message. - View Dependent Claims (20, 21, 22, 23, 24, 25, 26, 27, 28)
-
Specification