Silent training by error correction for on-line handwritting recognition systems

• US 5,544,260 A
• Filed: 07/12/1994
• Issued: 08/06/1996
• Est. Priority Date: 07/12/1994
• Status: Expired due to Fees

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Chefalas, Thomas E., Tappert, Charles C.
• Primary Examiner(s): Boudreau, Leo
• Assistant Examiner(s): KELLEY, CHRISTOPHER S
• Application Number:
  US08/274,074
• Time in Patent Office: 756 Days
• Field of Search: 382/13, 382/14, 382/15, 382/187, 382/189, 382/161
• US Class Current: 382/187
• CPC Class Codes: G06V 10/987 with the intervention of an operator

Abstract

A method for using information provided during error correction for modifying character prototypes in an on-line handwriting recognition system is disclosed. The method allows a user to correct misrecognized handwritten characters by overwriting directly on the displayed ASCII representation of the recognition result for a given character. The rewritten character is then used to silently retrain the system so as to adapt it to the user'"'"'s particular handwriting style.

Claims

1. A method of training a handwriting recognition system to recognize handwritten characters of a particular user, comprising:

(a) storing electrical representations of a plurality of prototype characters in a memory;

(b) entering a string of user-generated handwritten characters on an electronic digitizing tablet;

(c) generating an electrical representation of the string of handwritten characters;

(d) comparing the electrical representation of the user generated string of characters with the representations of the prototype characters stored in memory;

(e) displaying on the digitizing tablet, for each user generated handwritten character, a corresponding character, the corresponding character being a prototype character whose electrical representation most closely resembles the electrical representation of the particular user-generated handwritten character;

(f) entering on the digitizing tablet, for each incorrectly recognized user-input character, a user-generated handwritten correction character;

(g) comparing an electrical representation of the user generated handwritten correction character with the representations of the prototype characters stored in memory, excluding the electrical representation of the incorrectly returned character;

(h) displaying on the digitizing tablet, for each correction character, a corresponding correction character, the corresponding correction character of a particular user-generated handwritten correction character being a prototype character whose electrical representation most closely resembles the electrical representation of the particular user-generated handwritten correction character;

(i) repeating steps (f) through (h) if the corresponding correction character does not match the user generated correction character;

(j) comparing the electrical representation of the misrecognized user generated handwritten character and the user generated correction character and determining whether they resemble one another within a predetermined threshold and;

if so, retraining the system to recognize the misrecognized user generated character and the user generated correction character as the corresponding correction character;

if not, retraining the system to recognize the user generated correction character as the corresponding correction character.

2. The method of claim 1, wherein the retraining is done without request by the user.

3. The method of claim 1, wherein each of the user generated handwritten characters is mapped to its corresponding character, and wherein an electrical representation of the mapping is stored in system memory.

4. A system for recognizing handwritten characters, comprising:

an input device comprising means for transforming a user input string of characters into electrical representations thereof;

a character prototype store for storing electrical representations of prototype characters;

a character matcher comprising means for matching electrical representations of the user input character with the electrical representations of the prototype characters to recognize the user input characters;

means for displaying the results of the recognition of the user input characters on the input device;

means for user correction of an incorrect result of the recognition for each user input character, the means for correction comprising means for the user entering a corrected user input character on the input device, and means for comparing an electrical representation of the corrected user input character with electrical representations of the prototype characters, except for the incorrect prototype, and returning a corrected result character;

means for comparing the misrecognized user generated handwritten character and the corrected user input character and determining whether they resemble one another within a predetermined threshold, and;

if so, retraining the system to recognize the misrecognized user generated character and the corrected user input character as the corrected result character;

if not, retraining the system to recognize the corrected user input character as the corrected result character.

9. The system of claim 4, wherein the input device comprises a digitizing tablet.

11. The system of claim 4, further comprising means for writing the corrected character directly on the displayed representation of the result character.

5. A method of training a handwriting recognition system to recognize handwritten characters of a particular user, comprising:

(a) storing plurality of prototype characters in a memory;

(b) entering a string of user-generated handwritten characters on an input device;

(c) comparing the user generated string of characters with the prototype characters stored in memory;

(d) displaying, for each user generated handwritten character, a corresponding character, the corresponding character of a particular user-generated character being a prototype character most closely resembling the particular user-generated handwritten character;

(e) entering a user-generated handwritten correction character for each incorrectly recognized user-input character;

(f) comparing the user generated correction characters with the prototype characters stored in memory, excluding the incorrectly returned character;

(g) displaying a corresponding correction character for each user generated correction character, the corresponding correction character of a particular user-generated correction character being a prototype character most closely resembling the particular user-generated handwritten correction character;

(h) repeating steps (e) through (g) if the corresponding correction character does not match the user generated correction character;

(i) comparing the misrecognized user generated handwritten character and the user generated correction character and determining whether they resemble one another within a predetermined threshold, and;

if so, retraining the system to recognize the misrecognized user generated character and the user generated correction character as the corresponding correction character;

if not, retraining the system to recognize the user generated correction character as the corresponding correction character.

6. The method of claim 5, further comprising:

retraining the system to recognize the user generated correction character as the corresponding correction character.

7. The method of claim 5, wherein each of the user generated handwritten characters is mapped to its corresponding character, and wherein the mapping is stored in system memory.

8. The method of claim 5, wherein each user generated correction character is mapped to its corresponding correction character, and wherein the mapping is stored in system memory.

10. The method of claim 5, wherein the retraining is performed without request by the user.