Script recognition for ink notes

• US 7,929,769 B2
• Filed: 12/13/2005
• Issued: 04/19/2011
• Est. Priority Date: 12/13/2005
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Lin, Zhouchen, Wang, Xianfang, Wang, Jian
• Primary Examiner(s): Le; Vu
• Assistant Examiner(s): Allison; Andrae S
• Application Number:
  US11/301,761
• Publication Number:
  US 20070133877A1
• Time in Patent Office: 1,953 Days
• Field of Search: 382/160, 382/174, 382186-187, 382/203, 382229-230, 382/296, 382/298, 703/8, 703/45, 703/46, 703/52, 703/251, 703/255, 703/2
• 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 30/2455 Discrimination between machine-print, hand-print and cursive writing
• CPC Class Codes: G06V 30/36 Matching; Classification

Abstract

Computer-readable media having computer-executable instructions distinguish the script type of at least one portion of a writing input. At least one sub-word of a writing line of a handwritten document is identified and is processed to determine the associated writing style that includes a cursive writing style and a hand-printed writing style. The writing line is consequently associated with a script type. The script type of a writing line is determined from the script types of the sub-words in the writing line. When the number of sub-words having a first script type is greater than the number of sub-words having a second script type, the script type of the writing line is categorized as the first script type. In addition, a script analyzer determines a writing style of at least one sub-word and selects one of a plurality of neural networks to categorize the script type of a writing line.

Claims

1. A computer-readable medium excluding signals, the computer-readable medium having computer-executable instructions for performing steps comprising:

(a) determining whether a sub-word of a writing line corresponds to a cursive writing style;

(b) determining whether the sub-word corresponds to a hand-printed writing style; and

(c) in response to (a) and (b), classifying a script type of the writing line as a first script type when a first number of sub-words that are classified as the first script type exceeds a second number of sub-words that are classified as a second script type.

2. The computer-readable medium excluding signals, the computer-readable medium of claim 1 having computer-executable instructions to perform:

(d) identifying a direction of the writing line; and

(e) utilizing the direction of the writing line for classifying the script type of the writing line.

3. The computer-readable medium excluding signals, the computer-readable medium of claim 1 having computer-executable instructions to perform:

(d) preprocessing the writing line before determining a writing style of the writing line.

4. The computer-readable medium excluding signals, the computer-readable medium of claim 3 having computer-executable instructions to perform:

(d)(i) approximating a stroke of the writing line by a polyline.

5. The computer-readable medium excluding signals, the computer-readable medium of claim 1 having computer-executable instructions to perform:

(c)(i) utilizing at least one neural network to classify the script type from a plurality of script types.

6. The computer-readable medium excluding signals, the computer-readable medium of claim 5 having computer-executable instructions to perform:

(c)(i)(l) utilizing a first neural network when the sub-word corresponds to the cursive writing style; and

(c)(i)(2) utilizing a second neural network when the sub-word corresponds to the hand-printed writing style.

7. The computer-readable medium excluding signals, the computer-readable medium of claim 2, wherein the script type is selected from the group consisting of a Han script and a Roman script, having computer-executable instructions to perform:

(f) classifying the script type as the Han script when the direction of the writing line corresponds to a vertical direction.

8. The computer-readable medium excluding signals, the computer-readable medium of claim 1 having computer-executable instructions to perform:

(d) grouping at least one stroke to form the sub-word.

9. The computer-readable medium excluding signals, the computer-readable medium of claim 1, wherein the script type is selected from the group consisting of a Han script and a Roman script, having computer-executable instructions to perform:

(d) classifying the script type as the Han script when the sub-word contains a number of strokes exceeding a predetermined threshold.

10. The computer-readable medium excluding signals, the computer-readable medium of claim 4 having computer-executable instructions to perform:

(d)(i)(l) approximating the stroke by linear regression.

11. The computer-readable medium excluding signals, the computer-readable medium of claim 10 having computer-executable instructions to perform:

(d)(i)(2) merging adjacent segments of the stroke.

12. The computer-readable medium excluding signals, the computer-readable medium of claim 5 having computer-executable instructions to perform:

(d) training the at least one neural network to classify the script type from a plurality of previously classified text examples.

13. The computer-readable medium excluding signals, the computer-readable medium of claim 8 having computer-executable instructions to perform:

(d)(i) associating a first stroke and a second stroke with the sub-word when a distance between the first stroke and the second stroke is less than a predetermined distance.

14. The computer-readable medium excluding signals, the computer-readable medium of claim 1 having computer-executable instructions to perform:

(a)(i) sub-dividing a stroke of the sub-word;

(a)(ii) determining a direction change of the sub-word; and

(a)(iii) categorizing a writing style of the writing line as being the cursive writing style from the direction change.

19. The computer-readable medium excluding signals, the computer-readable medium of claim 5 wherein the at least one neural network classifies the script type from a plurality of script types via input comprising segments of strokes, the segments of strokes comprising representations of five parameters.

15. An apparatus that distinguishes a script type, comprising:

a plurality of neural networks;

a script analyzer receiving a writing input, determining a writing style of at least one sub-word of a writing line contained in the writing input, and selecting one of the plurality of neural networks based on the writing style; and

said one of the plurality of neural networks classifying the script type of the writing line to provide a script style indicator, wherein script type is classified as a first script type when a first number of sub-words that are classified as the first script type exceeds a second number of sub-words that are classified as a second script type.

16. The apparatus of claim 15, further comprising:

a preprocessor approximating a stroke of one of the at least one sub-word by at least one polyline.

17. The apparatus of claim 15, the plurality of neural networks comprising:

a cursive writing neural network processing the at least one sub-word to determine the script of the writing line when the writing style corresponds to a cursive writing style; and

a hand-printed writing neural network processing the at least one sub-word to determine the script of the writing line when the writing style corresponds to a hand-printed writing style.

18. A computer-readable medium excluding signals, the computer-readable medium having computer-executable instructions for performing steps comprising:

(a) determining a direction of a writing line;

(b) classifying a script type as a Han script when the direction is a vertical direction;

(c) determining whether a sub-word corresponds to a cursive writing style;

(d) when the sub-word corresponds to the cursive writing style, utilizing a first neural network to determine the script type of the sub-word;

(e) determining a number of strokes in the sub-word;

(f) classifying the script type of the sub-word as the Han script when the number of strokes exceeds a predetermined threshold;

(g) when the number of strokes is equal or less than the predetermined threshold, utilizing a second neural network to determine the script type of the sub-word; and

(h) repeating (c)-(g);

(i) classifying the script type of the writing line as the Han script when a first number of sub-words that are classified as the Han script exceeds a second number of sub-words that are classified as a Roman script; and

(j) classifying the script type of the writing line as the Roman script when the second number of sub-words that are classified as the Roman script exceeds the first number of sub-words that are classified as the Han script.

20. A process comprising:

receiving a writing input including a writing line;

determining a writing style of a sub-word of the writing line, the writing style corresponding to one of a cursive-writing style or a hand-printed writing style;

distinguishing a first number of sub-words of the writing line as a first script type of a plurality of script types and distinguishing a second number of sub-words of the writing line as a second script type of the plurality of script types;

classifying the writing line as the first script type when the first number of sub-words of the first script type exceeds the second number of sub-words of the second script type; and

storing the first script type as a classification of the writing line.