System for recognizing handwritten words of cursive script

• US 5,515,455 A
• Filed: 05/02/1994
• Issued: 05/07/1996
• Est. Priority Date: 09/02/1992
• Status: Expired due to Fees

• Current Assignee: The Research Foundation for The State University of New York (State University of New York)
• Original Assignee: The Research Foundation for The State University of New York (State University of New York)
• Inventors: Wang, Dacheng, Govindaraju, Venugopal, Srihari, Sargur
• Primary Examiner(s): Boudreau, Leo H.
• Assistant Examiner(s): ANDERSON, DAVID
• Application Number:
  US08/236,252
• Time in Patent Office: 736 Days
• Field of Search: 382/3, 382/9, 382/13, 382/16, 382/19, 382/25, 382/26, 382/28, 382/55
• US Class Current: 382/186
• CPC Class Codes: G06V 30/2268 using stroke segmentation

Abstract

A system for recognizing handwritten words of cursive script by extracting temporal information from static text is described. The temporal information consists of the shape, location and temporal sequence of the segments, or "primitives", constituting the word and is translated from bit maps images of the words into digital information format. By comparing the digital information with formatted digital information for a plurality of known words, as may be stored in computer memory, a list of words having a relatively high probability of matching the static text is generated, and used for word recognition by a process of selecting the most likely match.

Claims

1. A method for recognizing a static handwritten word of cursive script, comprising:

optically reading said word and forming a bit map of pixels representing said word;

skeletonizing said word within said bit map;

segmenting said skeletonized word into one or more primitives, said skeletonized word including a plurality of feature points and said primitives each comprising a continuous segment of said skeletonized word extending between an original feature point and a terminal feature point;

forming a sequence representing the order in which said primitives were written by ordering said primitives in succession beginning at the left side of said word; and

classifying said word by comparing said primitives and said sequence with stored primitives and their associated sequences for known words, wherein the step of forming a sequence comprises;

locating a primitive which is left-most in said word, examining said left-most primitive for the presence of one or more of said end points and designating said left-most primitive as a first primitive if it contains one or more of said end points;

if said left-most primitive does not contain one or more of said end points, examining a primitive connected with said left-most primitive for the presence of an end point, and designating said connected primitive as said first primitive if it contains an end point and designating said left-most primitive as said first primitive if said connected primitive does not contain an end point; and

ordinally designating as subsequent primitives each of said primitives which are connected which are connected with said first primitive and with said subsequent primitives.

2. The method according to claim 1, wherein the step of skeletonizing comprises:

thresholding said bit map, such that said pixels having an intensity above a predetermined level are assigned a binary value of one and all other of said pixels are assigned a binary value of zero, said thresholded bit map thereby containing at least one continuous line consisting of said pixels having a value of zero, said at least one line beginning at an original point and ending at a terminal point; and

thinning said at least one continuous line to form a line with a width of approximately one pixel.

3. The method according to claim 1 wherein said step of segmenting said skeletonized word into primitives comprises:

beginning at a first one of said feature points, tracing adjacent pixels until a second one of said feature points is encountered, said first and second feature points and said adjacent pixels thereby constituting a primitive; and

storing the location of said first and second feature points and the number and direction of pixels between said first and second feature points.

4. The method according to claim 3, wherein said adjacent pixels are pixels having an assigned value of 1.

5. The method according to claim 3 wherein said first one of said points is either a corner point or an end point and said second one of said points is either a corner point or an end point.

6. The method according to claim 3 further comprising:

if said second point was not a terminal point, repeating the preceding step until one of said primitives contains a feature point which is said terminal end point.

7. The method according to claim 1 wherein said classifying comprises:

determining the length and direction of each of said primitives; and

comparing said length, direction and sequence of said primitives for said word with primitives of words stored in a memory device and generating a list of words having a high probability of matching said word.

8. The method according to claim 7 wherein said step of determining the length and direction comprises:

counting the number of said pixels constituting said primitive, said primitive having an original feature point and a terminal feature point; and

measuring;

the angle formed with horizontal by an imaginary straight line between said original and terminal feature points,

the angle formed with horizontal by an imaginary straight line between said original feature point and a pixel adjacent said original feature point; and

the angle formed with horizontal by an imaginary straight line between said terminal feature point and a pixel adjacent said terminal feature point.

9. The method according to claim 1 wherein the feature points comprise end feature points, corner feature points, and intersection feature points.

10. The method according to claim 9 wherein the original feature point and the terminal feature point may be either an end feature point or a corner feature point.

11. The method according to claim 9 wherein said intersection feature points are disposed between original feature points and terminal feature points.

12. The method according to claim 9 wherein the word is initially traced between the original feature point and the terminal feature point of each primitive.

13. The method according to claim 12 wherein each primitive having one of the intersection feature points is traced in a first manner.

14. The method according to claim 13 wherein each primitive having another one of the intersection feature points is traced in a second manner.

15. The method according to claim 1 wherein the handwritten word of cursive script comprises two or more connected characters and one or more of said primitives extends between two of the characters.

16. Apparatus for recognizing a static handwritten word of cursive script, comprising;

means for reading said word and forming a bit map of pixels representing said word;

means for skeletonizing said word within said bit map;

means for segmenting said skeletonized word into one or more primitives, said skeletonized word including a plurality of feature points and said primitives each comprising a continuous segment of said skeletonized word extending between an original feature point and a terminal feature point;

means for forming a sequence representing the order in which said primitives were written by ordering said primitives in succession beginning at the left side of said word; and

means for classifying said word by comparing said primitives and said sequence with each of a plurality of stored primitives and their associated sequences for known words, wherein said means for forming a sequence comprises;

means for locating a primitive which is left-most in said word, examining said left-most primitive for the presence of one or more of said end points and designating said left-most primitive as a first primitive if it contains one or more of said end points;

means for examining a primitive connected with said left-most primitive for the presence of an end point if said left-most primitive does not contain one or more of said end points, and for designating said connected primitive as said first primitive if it contains an end point and designating said left-most primitive as said first primitive if said connected primitive does not contain an end point; and

means for ordinally designating as subsequent primitives each of said primitives which are connected with said first primitive and with said subsequent primitive.

17. The apparatus according to claim 16, comprising:

means for thresholding said bit map, such that said pixels having an intensity above a predetermined level are assigned a binary value of one and all other of said pixels are assigned a binary value of zero, said thresholded bit map thereby containing at least one continuous line consisting of said pixels having a value of zero, said at least one line beginning at an original point and ending at a terminal point; and

means for thinning said at least one continuous line to form a line with a width of approximately one pixel.

18. The apparatus according to claim 16, wherein said means for segmenting said skeletonized word comprises:

means for tracing adjacent pixels, beginning at a first one of said points, until a second one of said points is encountered, said first and second points and said adjacent pixels thereby constituting a primitive; and

means for storing the location of said first and second points and the number and direction of pixels between said first and second points.

19. The apparatus according to claim 16 wherein said classifying comprises:

means for determining the length and direction of each of said primitives; and

means for comparing said length, direction and sequence of said primitives for said word with primitives of words stored in a memory device and for generating a list of words having a high probability of matching said word.

20. The apparatus according to claim 19 wherein said means for determining the length and direction comprises:

means for counting the number of said pixels constituting said primitive said primitive having an original feature point and a terminal feature point; and

means for measuring;

the angle formed with horizontal by an imaginary straight line between original and terminal feature points,

the angle formed with horizontal by an imaginary straight line between said original feature point and a pixel adjacent said original feature point; and

the angle formed with horizontal by an imaginary straight line between said terminal feature point and a pixel adjacent said terminal feature point.

21. The apparatus according to claim 16 wherein said means for reading comprises means for reading a static image.

22. The apparatus according to claim 21 wherein said static image reading means is an optical character reading device.

23. The apparatus according to claim 16 wherein said means for reading comprises means for reading characters dynamically.

24. The apparatus according to claim 16 wherein the feature points comprise end feature points, corner feature points, and intersection feature points.

25. The apparatus according to claim 24 wherein the original feature point and the terminal feature point may be either an end feature point or a corner feature point.

26. The apparatus according to claim 24 wherein said intersection feature points are disposed between original feature points and terminal feature points.

27. The apparatus according to claim 24 wherein the word is initially traced between the original feature point and the terminal feature point of each primitive.

28. The apparatus according to claim 27 wherein each primitive having one of the intersection feature points is traced in a first manner.

29. The apparatus according to claim 28 wherein each primitive having another one of the intersection feature points is traced in a second manner.

30. The apparatus according to claim 16 wherein the handwritten word of cursive script comprises two or more connected characters and one or more of said primitives extends between two of the characters.

31. A system for reading and sorting mail, comprising:

means for optically reading an address from a mail piece, said address comprising at least one static word;

means for forming a bit map of pixels representing said word;

means for skeletonizing said word within said bit map;

means for segmenting said skeletonized word into one or more primitives, said skeletonized word including a plurality of feature points and said primitives each comprising a continuous segment of said skeletonized word extending between an original feature point and a terminal feature point and wherein said word comprises two or more connected characters and one or more of said primitives extends between two of the characters;

means for forming a sequence representing the order in which said primitives were written by ordering said primitives in succession beginning at the left side of said word;

means for classifying said word by comparing said primitives and said sequence with each of a plurality of stored primitives and their associated sequences for known words; and

means for directing said mail piece in response to the result of the classification of said word, wherein said means for forming a sequence comprises;

means for locating a primitive which is the left-most in said word, examining said left-most primitive for the presence of one or more of said end points and designating said left-most primitive as a first primitive if it contains one or more of said end points

means for examining a primitive connected with said left-most primitive for the presence of an end point if said left-most primitive does not contain one or more of said end points, and for designating said connected primitive as said first primitive if it contains an end point and designating said left-most primitive as said first primitive if said connected primitive does not contain an end point; and

means for ordinally designating as subsequent primitives each of said primitives which are connected with said first primitive and with said subsequent primitives.