Trellis based word decoder with reverse pass

• US 9,047,540 B2
• Filed: 03/14/2013
• Issued: 06/02/2015
• Est. Priority Date: 07/19/2012
• Status: Expired due to Fees

• Current Assignee: Qualcomm, Inc.
• Original Assignee: Qualcomm, Inc.
• Inventors: Baheti, Pawan Kumar, Barman, Kishor K., Krishna Kumar, Raj Kumar
• Primary Examiner(s): Do, Anh
• Application Number:
  US13/829,960
• Publication Number:
  US 20140023273A1
• Time in Patent Office: 810 Days
• Field of Search: 382/229, 382/177, 382/290, 382/181, 382/305, 382/209, 382/159, 382/182, 382/155, 382/226, 382/242, 382/233, 382/232, 382/310, 375/265, 375/262, 375/341, 714/795, 358/1.15, 358/3.28, 235/375, 235/494, 704/254, 704/256, 704/256.7
• US Class Current: 1/1
• CPC Class Codes: G06V 30/10 Character recognition
• CPC Class Codes: G06V 30/262 using context analysis, e.g. lexical, syntactic or semantic context

Abstract

Systems, apparatuses, and methods to relate images of words to a list of words are provided. A trellis based word decoder analyses a set of OCR characters and probabilities using a forward pass across a forward trellis and a reverse pass across a reverse trellis. Multiple paths may result, however, the most likely path from the trellises has the highest probability with valid links. A valid link is determined from the trellis by some dictionary word traversing the link. The most likely path is compared with a list of words to find the word closest to the most.

Claims

1. A method to relate images of words to a list of words in an optical character recognition (OCR) system, the method comprising:

receiving a plurality of OCR characters corresponding to an image of a word, wherein the plurality of OCR characters are from an OCR system;

determining a most likely (ML) path based on the plurality of OCR characters applied to a loaded forward trellis and a loaded reverse trellis, thereby forming a decoded word; and

displaying the decoded word.

2. The method of claim 1, wherein the list of words comprise a Devanagari list of words.

3. The method of claim 1, further comprising:

capturing and extracting the image of the word form a captured image to form an extracted word;

segmenting the extracted word into character images;

recognizing the plurality of OCR characters from the character images; and

providing the plurality of OCR characters.

4. The method of claim 1, wherein the plurality of OCR characters comprises a single decoded character for each character position.

5. The method of claim 1, wherein the plurality of OCR characters comprises multiple characters for each character position and a corresponding likelihood of each character for the character position.

6. The method of claim 1, further comprising pre-processing comprising:

accessing the list of words;

loading a forward trellis using the list of words to form the loaded forward trellis; and

loading a reverse trellis using the list of words to form the loaded reverse trellis.

7. The method of claim 6, further comprising forming a linking vector that indicates valid and invalid links into a node.

8. The method of claim 6, further comprising, for each node in the forward trellis and the reverse trellis having a previous character position entering the node, forming a linking vector that indicates valid and invalid links into the node.

9. The method of claim 6, further comprising, for each node in the forward trellis and the reverse trellis having a previous character position entering the node, forming a set of words indicating which words from the list of words have a path through the node.

10. The method of claim 6, wherein loading the reverse trellis comprises, for each word in the list of words:

selecting a word from the list of words; and

loading the reverse trellis at;

a first character position with a final character;

a second character position with a second to the final character; and

a third character position with a third to the final character.

11. The method of claim 1, further comprising receiving the loaded forward trellis and the loaded reverse trellis.

12. The method of claim 1, further comprising:

receiving the loaded forward trellis loaded with the list of words in a forward-position order; and

receiving the loaded reverse trellis loaded with the list of words in a reverse-position order.

13. The method of claim 1, further comprising converting the plurality of OCR characters from a Unicode to a character identifier (char ID) to form a plurality of char IDs, wherein determining the ML path based on the plurality of OCR characters comprises determining the ML path based on the plurality of char IDs.

14. The method of claim 1, wherein determining the ML path comprises:

determining at least one forward path through the loaded forward trellis;

determining at least one reverse path through the loaded reverse trellis; and

selecting, as the ML path, from the at least one forward path and the at least one reverse path, having a highest likelihood.

15. The method of claim 14, wherein the at least one forward path through the loaded forward trellis comprises multiple paths through the loaded forward trellis.

16. The method of claim 14, wherein determining the at least one forward path through the loaded forward trellis comprises:

determining a link from a previous character position to a current character position is invalid from the loaded forward trellis; and

discarding the link from the previous character position to the current character position based on the link being invalid in the loaded forward trellis.

17. The method of claim 14, wherein determining the at least one forward path through the loaded forward trellis comprises:

determining a link from a previous character position to a current character position is valid from the loaded forward trellis; and

saving the link from the previous character position to the current character position based on the link being valid in the loaded forward trellis.

18. The method of claim 14, wherein determining the at least one forward path through the loaded forward trellis comprises:

accessing a linking vector for a node in a current character position;

wherein the linking vector indicates valid and invalid paths into the node.

19. The method of claim 1, wherein an ML word formed along the ML path is not in the list of words, the method further comprising:

finding a word from the list of words having a minimum distance to the ML word; and

updating the ML word with the word from the list of words.

20. The method of claim 19, wherein finding the word from the list of words having the minimum distance to the ML word comprises finding the word from the list of words having a minimum Hamming distance to the ML word.

21. The method of claim 1, further comprising translating an ML word, formed along the ML path, from a first language to second language.

22. The method of claim 1, further comprising concatenating an ML word, formed along the ML path, with other words determined from their respective ML path.

23. A mobile device to relate images of words to a list of words in an optical character recognition (OCR) system, the mobile device comprising:

a camera configured to capture an image of a word;

a processor coupled to the camera, the processor comprising code to;

receive a plurality of OCR characters corresponding to the image of the word; and

determine a most likely (ML) path based on the plurality of OCR characters applied to a loaded forward trellis and a loaded reverse trellis, thereby forming a decoded word; and

a display coupled to the processor and configured to display the decoded word.

24. The mobile device of claim 23, wherein the list of words comprise a Devanagari list of words.

25. The mobile device of claim 23, wherein the plurality of OCR characters comprise a single character for each character position.

26. The mobile device of claim 23, wherein the plurality of OCR characters comprise multiple characters for each character position and a corresponding likelihood of each character for the character position.

27. The mobile device of claim 23, further comprising a linking vector that indicates valid and invalid links into a node.

28. The mobile device of claim 23, further comprising a set of words indicating which words from the list of words have a path through a node.

29. The mobile device of claim 23, further comprising the loaded forward trellis and the loaded reverse trellis.

30. The mobile device of claim 23, further comprising a minimum Hamming distance module to find a word from the list of words having a minimum Hamming distance to an ML word, formed along the ML path.

31. A mobile device to relate images of words to a list of words in an optical character recognition (OCR) system, the mobile device comprising:

means for receiving a plurality of OCR characters corresponding to an image of a word, wherein the plurality of OCR characters are from an OCR system;

means for determining a most likely (ML) path based on the plurality of OCR characters applied to a loaded forward trellis and a loaded reverse trellis, thereby forming a decoded word; and

means for displaying the decoded word.

32. The mobile device of claim 31, wherein the list of words comprise a Devanagari list of words.

33. The mobile device of claim 31, wherein the plurality of OCR characters comprise a single character for each character position.

34. The mobile device of claim 31, wherein the plurality of OCR characters comprise multiple characters for each character position and a corresponding likelihood of each character for the character position.

35. The mobile device of claim 31, further comprising a linking vector that indicates valid and invalid links into a node.

36. The mobile device of claim 31, further comprising a set of words indicating which words from the list of words have a path through a node.

37. The mobile device of claim 31, further comprising the loaded forward trellis and the loaded reverse trellis.

38. The mobile device of claim 31, further comprising a minimum Hamming distance module to find a word from the list of words having a minimum Hamming distance to an ML word, formed along the ML path.

39. A device to relate images of words to a list of words in an optical character recognition (OCR) system, the device comprising a processor and a memory wherein the memory includes software instructions to:

receive a plurality of OCR characters corresponding to an image of a word, wherein the plurality of OCR characters are from an OCR system;

determine a most likely (ML) path based on the plurality of OCR characters applied to a loaded forward trellis and a loaded reverse trellis, thereby forming a decoded word; and

display the decoded word.

40. A non-transitory computer-readable storage medium including program code stored thereon, for a method to relate images of words to a list of words in an optical character recognition (OCR) system, comprising program code to:

receive a plurality of OCR characters corresponding to an image of a word, wherein the plurality of OCR characters are from an OCR system;

determine a most likely (ML) path based on the plurality of OCR characters applied to a loaded forward trellis and a loaded reverse trellis, thereby forming a decoded word; and

display the decoded word.

41. The non-transitory computer-readable storage medium of claim 40, wherein the list of words comprise a Devanagari list of words.

42. The non-transitory computer-readable storage medium of claim 40, wherein the plurality of OCR characters comprise a single character for each character position.

43. The non-transitory computer-readable storage medium of claim 40, wherein the plurality of OCR characters comprise multiple characters for each character position and a corresponding likelihood of each character for the character position.

44. The non-transitory computer-readable storage medium of claim 40, further comprising program code to interpreted a linking vector that indicates valid and invalid links into a node.

45. The non-transitory computer-readable storage medium of claim 40, further comprising program code to form a set of words indicating which words from the list of words have a path through a node.

46. The non-transitory computer-readable storage medium of claim 40, further comprising program code to receive the loaded forward trellis and the loaded reverse trellis.

47. The non-transitory computer-readable storage medium of claim 40, further comprising program code to find a word from the list of words having a minimum Hamming distance to an ML word, formed along the ML path.

48. A mobile device to relate images of words to a list of words in an optical character recognition (OCR) system, the mobile device comprising:

means for accessing the list of words;

means for loading a forward trellis using the list of words to form a loaded forward trellis; and

means for loading a reverse trellis using the list of words to form a loaded reverse trellis.

49. The mobile device of claim 48, wherein the list of words comprise a Devanagari list of words.

50. The mobile device of claim 48, further comprising means for forming a linking vector that indicates valid and invalid links into a node.

51. The mobile device of claim 48, further comprising, for each node in the forward trellis and the reverse trellis having a previous character position entering the node, means for forming a linking vector that indicates valid and invalid links into a node.

52. The mobile device of claim 48, further comprising, for each node in the forward trellis and the reverse trellis having a previous character position entering the node, means for forming a set of words indicating which words from the list of words have a path through the node.