Method and apparatus for character recognition using stop words

• US 6,252,988 B1
• Filed: 07/09/1998
• Issued: 06/26/2001
• Est. Priority Date: 07/09/1998
• Status: Expired due to Fees

• Current Assignee: DMT GmbH Feinwerktechnische Komplettlosungen, Lucent Technologies, Inc. (Nokia Corporation)
• Original Assignee: Lucent Technologies, Inc. (Nokia Corporation)
• Inventors: Ho, Tin Kam
• Primary Examiner(s): Au, Amelia M.
• Assistant Examiner(s): Dastouri, Mehrdad
• Application Number:
  US09/112,549
• Time in Patent Office: 1,083 Days
• Field of Search: 382/159, 382/160, 382/161, 382/229, 382/224, 382/226, 382/227, 382/228, 707/500
• US Class Current: 382/229
• CPC Class Codes: G06F 18/214 Generating training patterns; Bootstrap methods, e.g. bagging or boosting
• 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

An adaptive OCR technique for character classification and recognition without the input and use of ground truth derived from the image itself. A set of so-called stop words are employed for classifying symbols, e.g., characters, from any image. The stop words are identified independent of any particular image and are used for classification purposes across any set of images of the same language, e.g., English. Advantageously, an adaptive OCR method is realized without the requirement of the selection and inputting of ground truth from each individual image to be recognized.

Claims

1. A method for training an image classifier, the method comprising:

identifying a plurality of stop words, each stop word being from a same language and having an associated definition in such language, the plurality of stop words being identified as a function of a linguistic model and the plurality of stop words having an expected recognition coverage level associated therewith, wherein the plurality of stop words is limited to the following stop words;

a, about, after, all, also, an, and, any, are, as, at, back, be, because, been, before, being, between, both, but, by, can, could, day, did, do, down, each, even, first, for, from, get, good, had, has, have, he, her, here, him, his, how, I, if, in, into, is, it, its, just, know, life, like, little, long, made, make, man, many, may, me, men, more, most, Mr., much, must, my, never, new, no, not, now, of, old, on, one, only, or, other, our, out, over, own, people, said, same, see, she, should, so, some, state, still, such, than, that, the, their, them, then, there, these, they, this, those, three, through, time, to, too, two, under, up, very, was, way, we, well, were, what, when, where, which, who, will, with, work, world, would, year, years, you, and your;

comparing the plurality of stop words to a plurality of individual words in an input image, each stop word and each individual word being treated as a separate symbol during the comparing;

identifying matches between particular ones of the stop words and particular ones of the individual words of the input image, wherein each particular stop word matches a same particular individual word throughout the input image, to form a plurality of recognized words;

segmenting the plurality of recognized words to form a plurality of character prototypes; and

training the image classifier using the plurality of character prototypes to recognize at least one character from the input image.

2. The method of claim 1 wherein the identifying matches operation includes:

constructing a stop word classifier using a decision forest; and

applying the stop word classifier to the input image.

3. The method of claim 2 wherein the segmenting the plurality of recognized words operation includes:

aligning particular ones of the recognized words; and

extracting a plurality of common character prototypes as a function of the aligning between the particular ones of the recognized words.

4. The method of claim 3 wherein the extracting the plurality of common character prototypes operation further comprises:

estimating a width for each character of the particular ones of the recognized words; and

shifting and matching at least one pair of recognized words using the estimated width for each character of the particular ones of the recognized words of the pair.

5. The method of claim 3 wherein the input image is a printed document.

6. The method of claim 5 wherein the image classifier is part of an optical character recognition system.

7. The method of claim 2 wherein the constructing the stop word classifier operation further comprises:

generating a plurality of synthetic words corresponding to a particular stop word; and

identifying a feature vector for the particular stop word as a function of the corresponding plurality of synthetic words, and using the feature vector to construct the stop word classifier.

8. The method of claim 7 wherein the feature vector is a combination of a binary subsamples vector, a pixel correlation vector, a vertical runs count vector, and a horizontal runs count vector.

9. An optical character recognition apparatus comprising:

a selector for selecting at least one input image from an image source, the input image having a plurality of individual words;

an image symbol generator for comparing a plurality of stop words to the input image, each stop word being from a same language and having an associated definition in such language, the plurality of stop words being identified as a function of a linguistic model and the plurality of stop words having an expected recognition coverage level associated therewith, identifying matches between particular ones of the stop words and particular ones of the individual words of the input image, wherein each particular stop word matches a same particular individual word throughout the input image, to form a plurality of matching words, segmenting the plurality of matching words to form a plurality of character prototypes, wherein the plurality of stop words is limited to the following stop words;

a, about, after, all, also, an, and, any, are, as, at, back, be, because, been, before, being, between, both, but, by, can, could, day, did, do, down, each, even, first, for, from, get, good, had, has, have, he, her, here, him, his, how, I, if, in, into, is, it, its, just, know, life, like, little, long, made, make, man, many, may, me, men, more, most, Mr., much, must, my, never, new, no, not, now, of, old, on, one, only, or, other, our, out, over, own, people, said, same, see, she, should, so, some, state, still, such, than, that, the, their, them, then, there, these, they, this, those, three, through, time, to, too, two, under, up, very, was, way, we, well, were, what, when, where, which, who, will, with, work, world, would, year, years, you, and your;

an image classifier for classifying at least one character from the input image using the plurality of character prototypes; and

an image recognizer for producing at least one recognized image from the image source using the at least one character.