Writing recognition unit

• US 6,125,208 A
• Filed: 09/10/1996
• Issued: 09/26/2000
• Est. Priority Date: 03/10/1994
• Status: Expired due to Fees

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Maier, Udo, Ruppert, Werner
• Primary Examiner(s): Au, Amelia
• Assistant Examiner(s): AHMED, SAMIR ANWAR
• Application Number:
  US08/704,565
• Time in Patent Office: 1,477 Days
• Field of Search: 382/187, 382/209, 382/213, 382/215, 382/216, 382/219, 382/226, 382/227, 382/228, 395/76
• US Class Current: 382/228
• CPC Class Codes: G06F 18/254 of classification results, e.g. of results related to same input data
• CPC Class Codes: G06V 30/194 References adjustable by an adaptive method, e.g. learning

Abstract

An apparatus and method for recognizing written characters. In one aspect, a method for recognizing written characters comprises the steps of: recognizing written characters with each of a plurality of writing recognition units, each of the writing recognition units supplying reliability values for each recognized character; determining a confidence level for each of the written characters from the corresponding reliability value for the written characters; testing the agreement of the results from the writing recognition units for each of the recognized characters; evaluating a plurality of rules for each of the recognized character with the corresponding levels and agreements; and delivering the overall result of the writing recognition process.

Claims

1. An apparatus for recognizing written characters, comprising:

at least two writing recognition units, each of said writing recognition units generating recognition results corresponding to each recognized character, wherein each of the recognition results from each unit comprises at least one recognized symbol and a corresponding reliability value;

means for designating a recognition output parameter to each recognized character based on the corresponding recognition results, wherein the recognition output parameter for a given recognized character is designated as either a first value, if the corresponding recognition results of the at least two writing recognition units are in agreement, or a second value, if the corresponding recognition results of the at least two recognition units are not in agreement;

means for allocating at least one confidence level from a set of confidence levels to each recognized character based on the corresponding reliability values, wherein each of the confidence levels is defined by a predetermined range of reliability values;

means for allocating at least one match value for each confidence level allocated to each recognized character, wherein the match value represents the degree to which the confidence level matches the corresponding reliability value of a given recognized character;

means for generating a sample vector that represents a reliability value and confidence level variable for each of the corresponding recognition results for a given recognized character, wherein the sample vector comprises a plurality of elements each corresponding to one reliability value at one of the defined confidence levels, and wherein the at least one match value is stored in the element of the sample vector that corresponds to that confidence level; and

evaluating means for evaluating the confidence levels and recognition output parameter associated with each recognized character using a plurality of rules to generate an overall recognition result for each recognized character based on output results of the rules.

3. The apparatus of claim 1, further comprising means for calculating a degree of fulfillment for each of the plurality of rules that produce an output results, wherein the degree of fulfillment for a given rule is computed based on the match values of the confidence levels that are evaluated with the given rule.

4. The apparatus of claim 1, further comprising means for assigning to each of the plurality of rules that produce an output result, a degree of fulfilment that is equal to the lowest value of the match values of the confidence levels that are evaluated with the given rule.

5. The apparatus of claim 3, further comprising a degree of fulfillment table for storing the degree of fulfilment and corresponding output result for each rule.

6. The apparatus of claim 3, wherein the evaluation means selects, as the overall recognition result for a recognized character, the output result of the rule having the highest degree of fulfillment.

7. The apparatus of claim 1, wherein the output result of each of the plurality of rules is a variable that is one of a value indicating that a particular one of the at least two recognition units provides the most acceptable recognized symbol according to the given rule and a value indicating that none of the at least two recognition units provides an acceptable recognized symbol according to the rule.

8. The apparatus of claim 1, wherein the sample vectors are used as input variables by the evaluation means.

2. A method for recognizing written characters, comprising the steps of:

recognizing written characters with at least two writing recognition units, each of said writing recognition units supplying recognition results corresponding to each recognized character, the recognition results comprising at least one recognized symbol and a corresponding reliability value;

allocating at least one confidence level for each of the recognized characters from a set of confidence levels based on the corresponding reliability values for said recognized characters, wherein each of the confidence levels is defined by a predetermined range of reliability values;

designating a recognition output parameter for each recognized character based on the corresponding recognition results, wherein the recognition output parameter for a given recognized character is designated as either a first value, if the corresponding recognition results of the at least two writing recognition units are in agreement, or a second value, if the corresponding recognition results of the at least two recognition units are not in agreement;

allocating at least one match value for each confidence level allocated to each recognized character, wherein the match value represents the degree to which the confidence level matches the corresponding reliability value of a given recognized character;

generating a sample vector that represents a reliability value and confidence level variable for each of the corresponding recognition results for a given recognized character, wherein the sample vector comprises a plurality of elements each corresponding to one reliability value at one of the defined confidence levels, and wherein the at least one match value is stored in the element of the sample vector that corresponds to that confidence level;

evaluating a plurality of rules for each of said recognized characters with the corresponding confidence levels and recognition output parameters; and

generating an overall recognition result for each recognized character based on output results of the rules.

9. The method of claim 2, further comprising the step of calculating a degree of fulfillment for each of the plurality of rules that produce an output result, wherein the degree of fulfillment for a given rule is computed based on the match values of the confidence levels that are evaluated with the given rule.

10. The method of claim 2, further comprising the step of assigning to each of the plurality of rules that produce an output result a degree of fulfilment that is equal to the lowest value of the match values of the confidence levels that are evaluated with the given rule.

11. The method of claim 9, wherein the step of evaluating comprises the step of selecting, as the overall recognition result for a recognized character, the output result of the rule having the highest degree of fulfillment.

12. A program storage device readable by a machine, tangibly embodying a program of instructions executable by the machine to perform method steps for recognizing written characters, the method comprising the steps of:

recognizing written characters with at least two writing recognition units, each of said writing recognition units supplying recognition results corresponding to each recognized character, the recognition results comprising at least one recognized symbol and a corresponding reliability value;

allocating at least one confidence level for each of the recognized characters from a set of confidence levels based on the corresponding reliability values for said recognized characters, wherein each of the confidence levels is defined by a predetermined range of reliability values;

designating a recognition output parameter for each recognized character based on the corresponding recognition results, wherein the recognition output parameter for a given recognized character is designated as either a first value, if the corresponding recognition results of the at least two writing recognition units are in agreement, or a second value, if the corresponding recognition results of the at least two recognition units are not in agreement;

allocating at least one match value for each confidence level allocated to each recognized character, wherein the match value represents the degree to which the confidence level matches the corresponding reliability value of a given recognized character;

generating a sample vector that represents a reliability value and confidence level variable for each of the corresponding recognition results for a given recognized character, wherein the sample vector comprises a plurality of elements each corresponding to one reliability value at one of the defined confidence levels, and wherein the at least one match value is stored in the element of the sample vector that corresponds to that confidence level;

evaluating a plurality of rules for each of said recognized character with the corresponding confidence levels and recognition output parameters; and

generating an overall recognition result for each recognized character based on output results of the rules.

13. The program storage device of claim 12, further comprising instructions for performing the step of calculating a degree of fulfillment for each of the plurality of rules that produce an output result, wherein the degree of fulfillment for a given rule is computed based on the match values of the confidence levels that are evaluated with the given rule.

14. The program storage device of claim 12, further comprising instructions for performing the step of assigning to each of the plurality of rules that produce an output result, a degree of fulfilment that is equal to the lowest value of the match values of the confidence levels that are evaluated with the given rule.

15. The program storage device of claim 13, wherein the instructions for performing the step of evaluating comprise instructions for performing the step of selecting, as the overall recognition result for a recognized character, the output result of the rule having the highest degree of fulfillment.