Multi-scale recognizer for hand drawn strokes

• US 5,038,382 A
• Filed: 10/13/1989
• Issued: 08/06/1991
• Est. Priority Date: 10/13/1989
• Status: Expired due to Fees

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Lipscomb, James S.
• Primary Examiner(s): Boudreau, Leo H.
• Assistant Examiner(s): Fox, David
• Application Number:
  US07/421,211
• Time in Patent Office: 662 Days
• Field of Search: 382/13, 382/21, 382/56
• US Class Current: 382/189
• CPC Class Codes: G06V 30/347 Sampling; Contour coding; Stroke extraction

Abstract

A known hand drawn stroke is serially sampled at different scales by n angle filters to provide n filtered known strokes at n different scales. An unknown hand drawn stroke is sampled in a like manner to provide n filtered unknown strokes at n different scales. Filtered unknown strokes are compared with filtered known strokes. If there is a comparison within predetermined limits at any scale, the unknown hand drawn stroke is recognized as being the same as the known hand drawn stroke.

Claims

1. A handwritten character recognition apparatus comprising:

means for sampling a known handwritten character;

n filters, where n is an integer ≧

2, with the first such filter filtering out direction changes less than a first angle in said known handwritten character to provide a first filtered known character, with the second such filter filtering out direction changes less than a second angle in said first filtered known character, where said second angle is greater than said first angle for providing a second filtered known character and so on, with the nth filter filtering out direction changes less than an nth angle in the (nth-1) filtered known character, where the nth angle is greater than the (nth-1) angle, to provide an nth filtered known character; and

storing the n filtered known characters for subsequent comparison to an unknown handwritten character.

2. The combination claimed in claim 1, including means in each of said n filters for removing lines in a filtered character that are shorter than a given length.

3. The combination claimed in claim 2, including means for removing any retrace from at least said first filtered known character.

4. The combination claimed in claim 1, including:

means for sampling an unknown handwritten character;

m filters, where m is an integer ≧

2, with the first such filter filtering out direction changes less than a first angle in said unknown handwritten character to provide a first filtered unknown character, with the second such filter filtering out direction changes less than a second angle in said first filtered unknown character, where said second angle is less than said first angle, for providing a second filtered unknown character and so on, with the mth filter filtering out direction changes less than an mth angle in the (mth-1) filtered unknown character, when the mth angle is greater than the (mth-1) angle, to provide an mth filtered unknown character;

comparison means for comparing said first filtered unknown character with said n filtered known characters, for comparing said second filtered unknown character with said n filtered known characters and so on, and for comparing said mth filtered unknown character with said n filtered known characters; and

means for recognizing said unknown character as said known character if any comparison matches within predetermined limits.

5. The combination claimed in claim 4, wherein each of said m filters includes means for removing lines in a filtered unknown character that are shorter than a given length.

6. The combination claimed in the claim 5, including means for removing any retrace from at least said first filtered unknown character.

7. A handwritten character recognition apparatus, comprising:

means for sampling a known handwritten character;

a first means for filtering out direction changes less than a first angle in said sampled known handwritten character to provide a first filtered known character;

at least a second means for filtering out direction changes less than a second angle, where said second angle is larger than said first angle, in said first filtered known character to provide a second filtered known character;

means for storing said first filtered known character and said second filtered known character;

means for sampling an unknown handwritten character;

a third means for filtering out direction changes less than said first angle in said sampled unknown handwritten character to provide a first filtered unknown character;

at least a fourth means for filtering out direction changes less than said second angle in said first filtered unknown character to provide a second filtered unknown character;

a first comparison means for comparing said first filtered unknown character with said stored n filtered known characters;

a second comparison means for comparing said second filtered unknown character with said stored n filtered known characters; and

means for recognizing said unknown character as said known character if either comparison by said first or second comparison means matches within predetermined limits.

8. A handwritten character recognition apparatus, comprising:

means for sampling a known handwritten character;

a first means for filtering out direction changes and less than a first angle and lines less than a first length in said sampled known handwritten character to provide a first filtered known character;

means for removing any retrace in said first filtered known character;

at least a second means for filtering out direction changes less than a second angle, where said second angle is larger than said first angle, and lines less than a second length, in said first filtered known character with any retrace removed to provide a second known character;

means for removing any retrace in said second filtered known character;

means for storing said first filtered known character and said second filtered known character with any retrace removed;

means for sampling an unknown handwritten character;

a third means for filtering out direction changes less than said first angle and lines less than a first length in said sampled unknown handwritten character to provide a first filtered unknown character;

means for removing any retrace in said first filtered unknown character;

at least a fourth means for filtering out direction changes less than said second angle and lines less than said second length in said first filtered unknown character to provided a second filtered unknown character;

means for removing any retrace in said second filtered unknown character;

a first comparison means for comparing said first filtered unknown character with retrace removed with said stored n filtered known characters with retrace removed;

a second comparison means for comparing said second filtered unknown character with retrace removed with said stored n filtered known characters with retrace removed; and

means for recognizing said unknown character as said known character if either comparison by said first or second comparison means matches within predetermined limits.

9. A method of recognizing hand drawn strokes made up of a plurality of line segments connecting xy points, said method comprising the steps of:

sampling a known hand drawn stroke;

filtering out all adjacent line segments in said known hand drawn stroke that form an angle less than a first angle at their common xy point, and replacing said adjacent line segments with a line segment between their uncommon xy points, to form a first filtered known stroke;

filtering out all adjacent line segments in said first filtered known stroke that form an angle less than a second angle at their common xy point, and replacing said adjacent line segments with a line segment between their uncommon xy points to form a second filtered known stroke; and

storing said first filtered known stroke and said second filtered known stroke for subsequent comparison to an unknown hand drawn stroke.

10. The method of claim 9, including the steps of:

removing any line segments in said first and second filtered known strokes that are less than a predetermined length.

11. The method of claim 10, including the step of:

removing any retrace in said first and second filtered known strokes.

12. The method of claim 9, including the steps of:

sampling an unknown hand drawn stroke;

filtering out all adjacent line segments in said unknown hand drawn stroke that forms an angle less than a first angle at their common xy point, and replacing said adjacent line segments with a line segment between their uncommon xy points, to form a first filtered unknown stroke;

filtering out all adjacent line segments in said first filtered unknown stroke that form an angle less than a second angle at their common xy point, and replacing said adjacent line segments with a line segment between their uncommon xy points to form a second unknown stroke;

comparing said first filtered unknown stroke with said n filtered known strokes;

comparing said second filtered unknown stroke with said n filtered known strokes; and

recognizing said unknown hand drawn stroke as said known hand drawn stroke if either comparison is within predetermined limits.

13. The method of claim 12, including the step of:

removing any line segments in said first and second filtered unknown strokes that are less than a predetermined length.

14. The method of claim 13, including the step of:

removing any retrace in said first and second filtered unknown strokes.