Method and apparatus for recognizing handwritten words

• US 5,903,668 A
• Filed: 06/23/1994
• Issued: 05/11/1999
• Est. Priority Date: 05/27/1992
• Status: Expired due to Term

• Current Assignee: Apple Inc.
• Original Assignee: Apple Computer Incorporated (Apple Inc.)
• Inventors: Beernink, Ernest H.
• Primary Examiner(s): Couso, Jose L.
• Application Number:
  US08/265,497
• Time in Patent Office: 1,783 Days
• Field of Search: 382/9, 382/10, 382/13, 382/22, 382/25, 382/30, 382/34, 382/48
• US Class Current: 382/187
• CPC Class Codes: G06V 30/1423 the instrument generating sequences of position coordinates corresponding to handwriting preprocessing or recognising digital ink G06V30/32
• CPC Class Codes: G06V 30/333 Preprocessing; Feature extraction
• CPC Class Codes: G06V 30/373 using a special pattern or subpattern alphabet

Abstract

A method for the angle-independent recognition of handwritten objects including: a) receiving a handwritten stroke; b) normalizing the stroke; c) matching the normalized stroke against a stroke database to obtain at least one character part interpretation; and d) recognizing a handwritten object using one or more of the character part interpretations. Preferably, the step of normalizing the stroke includes the step of normalizing an angle of the stroke by first determining a stroke angle and then rotating the stroke by the stroke angle until the stroke is at a normalized position. An apparatus for the angle independent recognition of handwritten objects includes: a) a central processing unit (CPU); b) a pointer mechanism coupled to the CPU for providing a handwritten stroke to the CPU; c) a mechanism for normalizing the strokes provided by the pointer mechanism; d) a mechanism for matching the normalized strokes against a stroke database to obtain a number of character part interpretations; and e) a mechanism for recognizing a handwritten object using the character part interpretations. Preferably, the mechanism for normalizing the strokes includes a mechanism for detecting stroke angles of the strokes, and a mechanism for rotating the strokes by the stroke angles to create normalized strokes. The mechanism for recognizing the handwritten object preferably includes a controller, a number of recognition domains coupled to the controller, and an arbiter coupled to the controller for resolving multiple interpretations of the object.

Claims

1. A method for recognizing handwritten entries on a display screen of a computer system, the computer system also having a processor and memory, the method comprising the steps of:

receiving a handwritten stroke written on the screen of said computer system, said stroke being received in a form that can be manipulated by said processor;

determining a start point and a stop point of said stroke;

normalizing said stroke with said processor by rotating said stroke about a stroke center until at least one of said start point or said stop point aligns with a predefined axis, thereby obtaining a normalized stroke;

matching said normalized stroke against a template of character parts stored in memory to create one or more associated character part interpretations;

using said processor to group said normalized stroke, if possible, with one or more normalized strokes that were matched in a like fashion to create a character part group; and

using said processor to recognize a higher level object from said character part group utilizing the angles between the strokes of the character part group and using said character part interpretations.

2. A method as recited in claim 1 wherein said higher level object is a character.

3. A method as recited in claim 1 wherein said higher level object is a word.

4. A method as recited in claim 3 wherein entries in said template of character parts have been pre-rotated to match the normalized position of said character part.

5. A method as recited in claim 3 wherein said step of using a processor to group said normalized strokes groups those strokes that are related in space and time.

6. A method as recited in claim 3 wherein said step of recognizing a higher level object comprises classifying said character part group to provide a plurality of weighted interpretations, and choosing the best weighted interpretation.

7. A method for the recognition of handwritten objects on a display screen of a computer system, the computer system also having a processor and a memory, the method comprising the steps of:

receiving a handwritten stroke written on the display screen of said computer system, said stroke being received in a form that can be manipulated by said processor;

obtaining a start point and a stop point of said stroke;

normalizing said stroke by rotating said stroke about a stroke center until at least one of said start point or said stop point aligns with a vertical axis, whereby a normalized stroke is obtained;

matching said normalized stroke against a stroke database stored in memory to obtain at least one character part interpretation; and

using said processor to recognize a handwritten object using one or more of said character part interpretations.

8. A method as recited in claim 7 further comprising the step of filtering said handwritten stroke before normalizing said stroke.

9. A method as recited in claim 7 wherein said step of normalizing said stroke comprises rotating said stroke by a predefined angle to create a normalized stroke.

10. A method as recited in claim 9 wherein said step of normalizing said stroke further comprises the step of normalizing an aspect ratio of said stroke.

11. A method as recited in claim 9 wherein at least one of said starting point and said ending point are at a predetermined position after said rotating step.

12. A method as recited in claim 9, further comprising the step of creating a stroke template database having a plurality of stroke template entries against which said normalized stroke may be matched by said matching step.

13. A method as recited in claim 12 wherein said stroke template entries have been normalized to match the normalization of said normalized stroke.

14. A method as recited in claim 13 further comprising the step of creating a character description database, and the step of creating pointers from said stroke template entries to said character description database.

15. A method as recited in claim 14 further comprising the step of creating a character part description database, and the step of creating pointers from said character description database to said character part description database.

16. A method as recited in claim 9 wherein said step of recognizing a handwritten object utilizes said stroke angles to help provide weights for multiple interpretations for said object.

17. An apparatus for the angle-independent recognition of handwritten objects comprising:

a display screen;

a central processing unit (CPU);

pointer means for writing a handwritten stroke on said display screen;

means for providing the handwritten stroke to said CPU;

means for demarcating a start point and a stop point of said stroke;

means for normalizing said stroke by rotating said stroke until one of said start point or said stop point aligns with a predefined axis;

means for matching said normalized strokes against a stroke database stored in memory to obtain a plurality of character part interpretations; and

means for recognizing, with said CPU, a handwritten object using said character part interpretations.

18. An apparatus as recited in claim 17 wherein said means for normalizing strokes comprises:

means for detecting stroke angles of said strokes relative to said predefined axis with said CPU; and

means for rotating said strokes with said CPU by said stroke angles to create normalized strokes.

19. An apparatus as recited in claim 18 wherein said means for normalizing strokes further comprises:

means for detecting the aspect ratios of said strokes; and

means for normalizing said aspect ratios of said strokes if said aspect ratios are not within predetermined limits.

20. An apparatus as recited in claim 18 wherein said means for matching said normalized strokes includes a stroke template database having a plurality of stroke template entries corresponding to a plurality of character parts.

21. An apparatus as recited in claim 20 wherein said plurality of stroke template entries are normalized to match the orientation of said plurality of strokes.

22. An apparatus as recited in claim 18 wherein said means for recognizing comprises:

controller means;

a plurality of recognition domains coupled to said controller means; and

an arbiter coupled to said controller means for resolving multiple interpretations of said object.

23. An apparatus as recited in claim 22 wherein said plurality of recognition domains includes a character part recognition domain and a word recognition domain.

24. A method for the recognition of handwritten objects on a display screen of a computer system, the computer system also having a processor and a memory, the method comprising the steps of:

receiving a handwritten stroke written on the display screen of said computer system, said stroke being received in a form that can be manipulated by said processor;

normalizing the stroke by adjusting the aspect ratio of a bounding box surrounding the stroke when the aspect ratio is greater than a first predetermined valve or less than a second predetermined value, said step of normalizing reducing recognition errors due to distortion;

normalizing said stroke by a process including the steps of;

(i) identifying the endpoints of the stroke

(ii) identifying two of the stroke'"'"'s most widely spaced points, at least one of which is an endpoint of the stroke,

(iii) defining a line between these points,

(iv) determining an angle between the defined axis and the line defined in step (iii), and

(v) rotating the stroke by the angle determined in step (iv) such that at least one of the stroke'"'"'s endpoints lies on said defined axis;

matching the normalized stroke against a stroke database stored in memory to obtain at least one character part interpretation; and

using the processor to recognize a handwritten object using one or more of the character part interpretations.

25. A method as recited in claim 24 wherein the first predetermined value is about 3 and the second predetermined value is about one-third.

26. A method as recited in claim 24 wherein the step of normalizing said stroke further comprises rotating the stroke by a predefined angle.

27. A method for recognizing handwritten entries on a display screen of a computer system, the computer system also having a processor and memory, the method comprising the following steps:

receiving a handwritten stroke written on the display screen of said computer system, said stroke being received in a form that can be manipulated by said processor;

normalizing said stroke by a process including the steps of;

(i) identifying the endpoints of the stroke

(ii) identifying two of the stroke'"'"'s most widely spaced points, at least one of which is an endpoint of the stroke,

(iii) defining a line between these points,

(iv) determining an angle between the defined axis and the line defined in step (iii), and

(v) rotating the stroke by the angle determined in step (iv) such that at least one of the stroke'"'"'s endpoints lies on said defined axis;

matching said normalized stroke against a stroke database stored in memory to obtain at least one character part interpretation; and

using said processor to recognize a handwritten object using one or more of said character part interpretations, wherein said step of normalizing reduces recognition errors due to handwriting written at various angles.

28. The method or claim 27 further comprising a step of identifying the center of said stroke.

29. The method of claim 28 wherein the step of rotating the stroke by said angle comprises rotating the stroke about said center of the stroke.