Hidden Markov Model Based Handwriting/Calligraphy Generation

• US 20090041354A1
• Filed: 08/10/2007
• Published: 02/12/2009
• Est. Priority Date: 08/10/2007
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Liu, Peng, Wu, Yi-Jian, Ma, Lei, Soong, Frank Kao-PingK
• Granted Patent: US 7,983,478 B2
• Time in Patent Office: Days
• US Class Current: 382/187
• CPC Class Codes: G06F 18/295 Markov models or related models, e.g. semi-Markov models; Markov random fields; Networks embedding Markov models
• CPC Class Codes: G06V 30/287 of Kanji, Hiragana or Katakana characters

Abstract

An exemplary method for handwritten character generation includes receiving one or more characters and, for the one or more received characters, generating handwritten characters using Hidden Markov Models trained for generating handwritten characters. In such a method the trained Hidden Markov Models can be adapted using a technique such as a maximum a posterior technique, a maximum likelihood linear regression technique or an Eigen-space technique.

Claims

1. A method for handwritten character generation, implemented at least in part by a computing device, the method comprising:

receiving a character; and

generating a corresponding handwritten character using Hidden Markov Models trained to generate handwritten characters.

2. The method of claim 1 comprising receiving more than one character and generating corresponding handwritten characters.

3. The method of claim 1 wherein the Hidden Markov Models comprise Hidden Markov Models trained using slope, curvature and imaginary stroke features.

4. The method of claim 1 wherein the Hidden Markov Models comprise Hidden Markov Models trained using a real stroke feature set that comprises a slope feature and a curvature feature and an imaginary stroke feature set of a lesser dimension that comprises a slope feature.

5. The method of claim 1 wherein the generating further comprises using a Multi-Space Probability Distribution technique wherein the Multi-Space Probability Distribution technique includes a realness probability for determining whether a stroke is real or imaginary.

6. The method of claim 1 wherein the Hidden Markov Models comprise Hidden Markov Models trained using calligraphy data.

7. The method of claim 1 wherein the Hidden Markov Models comprise Hidden Markov Models trained using Eastern Asian character data.

8. The method of claim 1 wherein the Hidden Markov Models comprise Hidden Markov Models trained using data from a standard writer.

9. The method of claim 1 further comprising rendering the generated corresponding handwritten character using a pen model.

10. The method of claim 9 wherein the pen model comprises pen tip parameters.

11. The method of claim 1 wherein the Hidden Markov Models comprise Hidden Markov Models trained using a pressure feature that corresponds to writing pressure.

12. The method of claim 1 wherein the Hidden Markov Models comprise adapted, trained Hidden Markov Models, adapted using a technique selected from the group consisting of maximum a posterior techniques, maximum likelihood linear regression techniques and Eigen-space techniques.

13. The method of claim 1 wherein the Hidden Markov Models comprise adapted, trained Hidden Markov Models, adapted using a maximum likelihood linear regression techniques and ink data from the same writer as the received one or more characters.

14. The method of claim 1 wherein the trained Hidden Markov Models comprise adapted, trained Hidden Markov Models, adapted using an adaptation technique and ink data from a celebrity.

15. A method for adapting trained Hidden Markov Models for generation of handwritten characters, implemented at least in part by a computing device, the method comprising:

providing initial, trained Hidden Markov Models for generation of handwritten characters;

providing training ink data to adapt the initial, trained Hidden Markov Models; and

applying an adaptation technique to adapt the initial, trained Hidden Markov Models to the training ink data wherein the adaptation technique comprises a technique selected from a group consisting of maximum a posterior techniques, maximum likelihood linear regression techniques and Eigen-space techniques.

16. The method of claim 15 further comprising calculating feature vectors for the training ink data.

17. The method of claim 15 wherein the applying applies the selected technique to one or more mean vectors and one or more covariance matrix(ces).

18. The method of claim 15 wherein the initial, trained Hidden Markov Models comprise Hidden Markov Models trained using calligraphy data.

19. The method of claim 15 wherein the initial, trained Hidden Markov Models comprise Hidden Markov Models trained using a standard writer and the training ink data comprises data from a different writer.

20. A computing device for generating handwritten characters, the device comprising:

a processor;

a user input mechanism;

a display; and

control logic implemented at least in part by the processor to generate handwritten characters based on an algorithm that uses Hidden Markov Models.