Character recognition system and method

• US 7,327,883 B2
• Filed: 03/11/2003
• Issued: 02/05/2008
• Est. Priority Date: 03/11/2002
• Status: Active Grant

• Current Assignee: IMDS Software, Inc.
• Original Assignee: IMDS Software, Inc.
• Inventors: Polonowski, Jean-Pierre
• Primary Examiner(s): Mehta; Bhavesh M
• Assistant Examiner(s): Strege; John B
• Application Number:
  US10/384,649
• Publication Number:
  US 20030169925A1
• Time in Patent Office: 1,792 Days
• Field of Search: 382/182, 382/185, 382/186
• US Class Current: 382/182
• CPC Class Codes: G06F 18/24137 Distances to cluster centroïds

Abstract

A system and method for translating a written document into a computer readable document by recognizing the character written on the document aim at recognizing typed or printed, especially hand-printed or handwritten characters, in the various fields of a form. Providing a pixel representation of the written document, the method allows translating a written document into a computer readable document by i) identifying at least one field into the pixel representation of the document; ii) segmenting each field so as to yield at least one segmented symbol; iii) applying a character recognition method on each segmented symbol; and iii) assigning a computer-readable code to each recognized character resulting from the character recognition method. The character recognition method includes doing a vector quantization on each segmented symbol, and doing a vector classification using a vector base. A learning base is also created based on the optimal elliptic separation method. System and method according to the present invention allow to achieve a substitution rate of near zero.

Claims

1. A method for translating a written document into a computer-readable document, said method comprising:

providing a pixel representation of the written document;

providing a vector base comprising a plurality of classes, each said class corresponding to a computer-readable code and comprising a predetermined vector quantization of said corresponding computer-readable code;

identifying a field into said pixel representation of the written document;

segmenting said field, thereby yielding a segmented symbol;

performing a vector quantization on said segmented symbol;

comparing said vector quantization of said segmented symbol with said predetermined vector quantization of each said class, thereby assigning a similarity score between said segmented symbol and each said class; and

determining if only one of said similarity scores exceeds a predetermined threshold;

wherein when only one said similarity score exceeds said predetermined threshold, assigning to said segmented symbol a said computer-readable code corresponding to a said class for which said only one similarity score exceeds said predetermined threshold.

2. A method as recited in claim 1, wherein at least two said fields are identified on said pixel representation of the written document;

at least two recognition engines, each including a vector quantization method and a vector classification method, being used on applying a character recognition method on each said segmented symbol on said at least two fields.

3. A method as recited in claim 2, wherein a voting method is further used to allow the combination of the results of the at least two recognition engines using at least one of simple averages, weighted averages and veto rules.

4. A method as recited in claim 1, wherein comparing said vector quantization of said segmented symbol with said predetermined vector quantization of each said class is performed by using a separation method selected from the group consisting of an optimal linear separation method, an optimal elliptical separation method, a support vector machine, a quadratic vector machine, a modified quadratic discriminant function, and a K nearest neighbor.

5. A method as recited in claim 1, further comprising:

providing a pixel representation of the written document by digitizing said document.

6. A method as recited in claim 1, wherein said field is a generally rectangular portion of said pixel representation of the written document.

7. A method as recited in claim 1, further comprising:

recognizing the type of the written document;

associating a predetermined format to the written document corresponding to said type of the written document;

wherein, identifying said field into said pixel representation of the document is performed by using said predetermined format.

8. A method as recited in claim 7, wherein recognizing the type of the written document includes identifying a predetermined characteristic of the written document.

9. A method as recited in claim 8, wherein said predetermined characteristic includes a general configuration of the written document, text block, dimensions, and lines configurations.

10. A method as recited in claim 1, wherein assigning to said segmented symbol said computer-readable code further comprises:

quality controlling said segmented symbol assigned to said computer-readable code.

11. A method as recited in claim 1, further comprising an operator verifying the validity of segmented symbol assigned to a computer-readable code by comparing said segmented symbol assigned to said computer-readable code with a corresponding character on the written document.

12. A method as recited in claim 1, further comprising:

formatting said assigned computer-readable code.

13. A method as recited in claim 1, wherein said computer-readable code is an ASCII code.

26. A method as recited in claim 1, wherein when more than one said similarity score exceeds said predetermined threshold, said method further comprises submitting said vector quantization of said segmented symbol to a further inspection step.

27. A method as recited in any of claims 26, wherein said further inspection step includes at least one of (a) performing a human inspection step, (b) applying a predetermined field validity rule, (c) verifying in a stored database, (d) verifying in a thesaurus, and (e) applying an Intelligence Character Recognition method on said segmented symbol and a neighbour character segmented symbol.

28. A method as recited in claim 1, wherein when none of said similarity scores exceeds said predetermined threshold, said method further comprises submitting said vector quantization of said segmented symbol to a further inspection step.

14. A method for recognizing a character corresponding to a written symbol, said method comprising:

providing a pixel representation of the written symbol;

segmenting said pixel representation, yielding a segmented symbol;

doing a vector quantization on said segmented symbol, yielding a vector representation of said written symbol;

for possible class(i), i ranging from 1 to N, N being the number of different possible classes;

providing a vector representation(i) for each class(i);

computing a similarity score(i) using said vector representation (i) of said symbol and said vector representation for class(i); and

comparing said similarity score(i) to a threshold(i); and

if only one of said similarity score(x) is superior than the corresponding threshold(x), x ranging form 1 to N;

assigning to said written symbol a computer-readable code corresponding to said class(x).

15. A method as recited in claim 14, wherein said vector representation(i) for each class(i) is obtained through an optimal elliptical separation method.

16. A method as recited in claim 14, wherein said threshold(i) is computed using a vector base.

17. A method as recited in claim 14, wherein computing a similarity score(i) is done using a separation method selected from the group consisting of an optimal linear separation method, an optimal spherical separation method, a support vector machine, a quadratic vector machine, a modified quadratic discriminant function, and a K nearest neighbor.

18. A method as recited in claim 14, wherein said computer readable code is an ASCII code.

19. A system for translating a written document into a computer-readable document, said system comprising:

a document digitizer for creating a pixel representation of the document;

a controller coupled to said digitizer for;

receiving said pixel representation of the written document;

identifying a field in said pixel representation of the written document;

segmenting said field, thereby yielding a segmented symbol;

performing a vector quantization on said segmented symbol;

comparing said vector quantization of said segmented symbol with said predetermined vector quantization of each said class, thereby assigning a similarity score between said segmented symbol and each said class; and

determining if only one of said similarity scores exceeds a predetermined threshold;

wherein, when only one said similarity score exceeds said predetermined threshold, assigning to said segmented symbol a said computer-readable code corresponding to a said class for which said only one similarity score exceeds said predetermined threshold;

an output device coupled to said controller for displaying said segmented symbol; and

at least one input device coupled to said controller for entering a computer-readable code of a humanly recognized character among displayed segmented symbols.

20. A system as recited in claim 19, wherein said document digitizer is a scanner.

21. A system as recited in claim 19, wherein said document digitizer includes an optical reader.

22. A system as recited in claim 19, wherein said controller is selected from the group consisting of a computer, an electronic control circuit and a programmed chip.

23. A system as recited in claim 19, further comprising a storing device.

24. A system as recited in claim 19, wherein said at least one input device includes at least one of a pointing device and a keyboard.

25. A system for translating a written document into a computer readable document, said system comprising:

means for providing a pixel representation of the written document;

means for providing a vector base comprising a plurality of classes, each said class corresponding to a computer-readable code and comprising a predetermined vector quantization of said corresponding computer-readable code;

means for identifying a field into said pixel representation of the written document;

means for segmenting said field, thereby yielding a segmented symbol;

means for performing a vector guantization on said segmented symbol;

means for comparing said vector quantization of said segmented symbol with said predetermined vector quantization of each said class, thereby assigning a similarity score between said segmented symbol and each said class; and

means for determining if only one of said similarity scores exceeds a predetermined threshold;

wherein when only one said similarity score exceeds said predetermined threshold, assigning to said segmented symbol a said computer-readable code corresponding to a said class for which said only one similarity score exceeds said predetermined threshold.

29. A method as recited in 28, wherein said further inspection step includes at least one of (a) performing a human inspection step, (b) applying a predetermined field validity rule, (c) verifying in a stored database, (d)verifying in a thesaurus, and (e) applying an Intelligence Character Recognition method on said segmented symbol and a neighbour character segmented symbol.