Interleaved segmental method for handwriting recognition

• US 5,878,164 A
• Filed: 10/08/1997
• Issued: 03/02/1999
• Est. Priority Date: 01/21/1994
• Status: Expired due to Fees

• Current Assignee: Lucent Technologies, Inc. (Nokia Corporation)
• Original Assignee: Lucent Technologies, Inc. (Nokia Corporation)
• Inventors: Turin, William, Brown, Michael Kenneth, Hu, Jianying
• Primary Examiner(s): Boudreau, Leo H
• Assistant Examiner(s): Tadayon, Bijan
• Application Number:
  US08/946,876
• Time in Patent Office: 510 Days
• Field of Search: 382/155, 382/156, 382/157, 382/158, 382/159, 382/160, 382/161, 382/173, 382/176, 382/177, 382/178, 382/179, 382/193, 382/217, 382/224, 382/225, 382/226, 382/227, 382/228 , 382/229, 382/230, 382/231, 382/312, 382/313, 382/314, 382/315, 382/301, 382/317, 382/321, 382/216, 382/218, 382/290, 382/181, 382/182, 382/200, 382/183, 382/184 , 382/201, 382/185, 382/186, 382/202, 382/187, 382/188, 382/207, 382/189, 382/190, 382/209, 382/191, 382/192, 382/208, 382/194, 382/195, 382/204, 382/197, 382/198 , 382/199, 382/203, 395/3, 395/10, 395/50, 395/20, 395/21, 395/60
• US Class Current: 382/190
• CPC Class Codes: G06V 30/2268 using stroke segmentation
• CPC Class Codes: G10L 15/18 using natural language modelling

Abstract

A method of handwriting recognition is provided that combines the efficiency of a point oriented system and the shape information of a segment oriented system in an HMM based handwriting recognition system. The partial segmentation hypotheses obtained using the point oriented features in a conventional Viterbi search are augmented with scores based on segmental shape measurements made on the hypothesized segments. In addition, a method for segmentation of a handwriting sample is accomplished by incorporating a feature to distinguish between continuous and not continuous strokes and influencing the indicated hypothesis score by adding a penalty to the contraindicated hypothesis score.

Claims

1. An improved method for performing handwriting recognition of a handwriting sample represented by a signal sequence, utilizing one or more point oriented feature signals to hypothesize a segment of said handwriting sample giving rise to a point oriented hypotheses score wherein the improvement comprises the steps of:

generating one or more segmental feature signals of said segment of said handwriting sample giving rise to a segmental hypothesis score;

augmenting said point oriented hypothesis score with a segmental matching score computed using said segmental hypothesis score; and

recognizing said handwriting sample based on said augmented point oriented hypothesis score.

2. A method according to claim 1 wherein said handwriting recognition system is a Hidden Markov Model based handwriting recognition system.

3. A method according to claim 1 wherein one of said segmental feature signals is based on an inter-segmental distance measure correlation metric.

4. A method according to claim 3 further comprising the step of:

preprocessing said handwriting sample so that it contains a predetermined number of points.

5. An improved method for recognizing a handwriting sample on the basis of a set of alternative hypothesis scores, said alternative hypothesis scores being accumulated at each of a set of selected hypothesis scoring points falling within said handwriting sample, each alternative hypothesis score of said set of alternative hypothesis scores representing a likelihood indication of an alternate model of said handwriting sample, said handwriting sample having one or more pen-up instances, each pen-up instance having associated therewith a spatial distance between a pen-lift point and a pen down point, said set of selected hypothesis scoring points including said pen-lift points and said pen down points, the improvement comprising the steps of:

detecting each of said pen-up instances within said handwriting sample;

calculating said spatial distance associated with each of said pen-up instances, respectively;

associating with each of said pen-up instances a predetermined gap feature signal indicative of whether said spatial distances are at least equal to, or less than, a predetermined gap threshold spatial distance value;

adjusting said set of accumulated hypothesis scores by adding a predetermined penalty to each of said accumulated hypothesis scores of said set of alternative hypothesis scores corresponding to said pen down point, indicating a model of said handwriting sample continuous between said pen lift point and said pen down point, where said predetermined gap feature signal indicates a spatial distance equal to or greater than said threshold value and by adding a predetermined penalty to each of said accumulated hypothesis scores of said set of alternative hypothesis scores corresponding to said pen down point, indicating a model of said handwriting sample not continuous between said pen lift point and said pen down point, where said predetermined gap feature signal indicates a spatial distance less than said gap threshold value; and

recognizing said handwriting sample based on said adjusted alternative hypothesis scores.

6. A method according to claim 5 wherein said predetermined gap feature signal is a binary feature signal.

7. A method according to claim 5 wherein said handwriting recognition system is a Hidden Markov Model based handwriting recognition system.

8. An apparatus for performing handwriting recognition of a handwriting sample represented by a signal sequence, utilizing point oriented feature signals to hypothesize a segment of said handwriting sample giving rise to a point oriented hypothesis score, the improvement comprising:

means for generating one or more segmental feature signals of said segment of said handwriting sample giving rise to a segmental hypothesis score;

means for augmenting said point oriented hypothesis score with a segmental matching score computed using said segmental hypothesis score; and

means for recognizing said handwriting sample based on said augmented point oriented hypothesis score.

9. A method for performing handwriting recognition of a handwriting sample comprising a string of interconnected symbols represented by a signal sequence, comprising the steps of:

utilizing one or more point oriented feature signals to hypothesize one or more segmentations of said string of interconnected symbols, each of said hypothesized segmentations having associated therewith a point oriented hypothesis score;

generating one or more segmental feature signals of each of said hypothesized segmentations giving rise to a segmental hypothesis score for each of said hypothesized segmentations;

respectively adding to said point oriented hypothesis score of each of said hypothesized segmentations a segmental matching score computed using said segmental hypothesis score of each of said hypothesized segmentations, resulting in an augmented hypothesis score for each of said hypothesized segments; and

recognizing said handwriting sample based on said augmented hypothesis score.

10. An apparatus for performing handwriting recognition of a handwriting sample comprising a string of interconnected symbols represented by a signal sequence, comprising:

means for hypothesizing one or more segmentations of said string of interconnected symbols utilizing one or more point oriented feature signals and generating a point oriented hypothesis score for each of said hypothesized segmentations;

means for generating one or more segmental feature signals of each of said hypothesized segmentations giving rise to a segmental hypothesis score for each of said hypothesized segments;

means for respectively adding to said point oriented hypothesis score for each of said hypothesized segmentations a segmental matching score computed using said segmental hypothesis score for each of said hypothesized segmentations resulting in an augmented hypothesis score for each of said hypothesized segmentations; and

means for recognizing said handwriting sample based on said augmented hypothesis score.