Systems and methods for adaptive handwriting recognition

• US 20040234128A1
• Filed: 05/21/2003
• Published: 11/25/2004
• Est. Priority Date: 05/21/2003
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Meek, Christopher A., Thiesson, Bo
• Granted Patent: US 7,184,591 B2
• Time in Patent Office: Days
• US Class Current: 382/186
• CPC Class Codes: G06F 18/254 of classification results, e.g. of results related to same input data
• 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 30/1918 Fusion techniques, i.e. combining data from various sources, e.g. sensor fusion

Abstract

The present invention utilizes generic and user-specific features of handwriting samples to provide adaptive handwriting recognition with a minimum level of user-specific enrollment data. By allowing generic and user-specific classifiers to facilitate in a recognition process, the features of a specific user'"'"'s handwriting can be exploited to quickly ascertain characteristics of handwriting characters not yet entered by the user. Thus, new characters can be recognized without requiring a user to first enter that character as enrollment or “training” data. In one instance of the present invention, processing of generic features is accomplished by a generic classifier trained on multiple users. In another instance of the present invention, a user-specific classifier is employed to modify a generic classifier'"'"'s classification as required to provide user-specific handwriting recognition.

Claims

1. A system that facilitates adaptive handwriting recognition, comprising:

a component that receives handwriting data; and

a recognition component that analyzes the handwriting data in connection with recognition thereof, the recognition component employing at least one generic classifier and at least one user-specific classifier in connection with the analysis.

2. The system of claim 1, at least one generic classifier applied against generic features of the handwriting data that exhibit substantial commonality across a plurality of users.

3. The system of claim 2, the generic features including, at least in part, at least one selected from the group consisting of scaled shape features, size features, and position features.

4. The system of claim 1, at least one user-specific classifier applied against specific features of the handwriting data that exhibit substantial uniqueness to a particular user.

5. The system of claim 4, the specific features including, at least in part, at least one selected from the group consisting of size features and position features.

6. The system of claim 1, further comprising a plurality of user specific classifiers, each of the respective user-specific classifiers being associated with respective sets of features.

7. The system of claim 6, the sets of features being disjoint.

8. The system of claim 1, the recognition component employing confusion rules that associates a class/character with a set of classes/characters that can be confusable for the generic classifier.

9. The system of claim 8, at least one user-specific classifier can overrule the generic classifier.

10. The system of claim 7, employing density models constructed for respective characters of a confusion rule.

11. The system of claim 10, at least one density model comprising an MLLR adapted density model.

12. The system of claim 7, employing a direct discriminative model constructed for respective characters of a confusion rule.

13. The system of claim 1, the generic classifier being an adapted version of an initially supplied generic classifier.

31. A computer readable medium having stored thereon computer executable components of the system of claim 1.

33. A device employing the system of claim 1 comprising at least one selected from the group consisting of a computer, a server, and a handheld electronic device.

14. A method for building a handwriting recognition system, comprising:

building at least one generic classifier utilizing handwriting data associated with a plurality of users; and

building at least one user specific classifier utilizing handwriting data supplied by at least one selected from the group consisting of a specific user and both a specific user and a plurality of users.

15. The method of claim 14, further comprising employing a generic classification model and at least one tying class having at least one tying set to train at least one user-specific classifier.

16. The method of claim 15, wherein data associated with at least one tying set for at least one tying class is utilized to adapt a set of models using MLLR.

17. The method of claim 15, wherein data associated with at least one tying set for at least one tying class is utilized to build a set of direct density models.

18. The method of claim 15, wherein data associated with at least one tying set for at least one tying class is utilized to build a direct discriminative model.

19. The method of claim 18, wherein the direct discriminative model is based on classifiers with class/character data tied to at least one finest partition of at least one tying set for at least one user-specific feature.

20. A system for recognizing handwriting, comprising:

means for generically determining a classification of handwriting information trained via data from multiple users; and

means for specifically determining a classification of handwriting information trained, at least in part, via data from a specific user; and

means for operably utilizing both the generically determining means and the specifically determining means to classify handwriting data.

21. The system of claim 20, further including:

means for classifying handwriting data for at least one generic classifier confusion rule.

22. The system of claim 21, further including:

means for overruling at least one generic classification based upon at least one confusion rule.

23. A method for recognizing handwriting, comprising:

processing handwriting data utilizing at least one generic classifier to establish at least one classification of the data; and

utilizing at least one user-specific classifier to further refine the classification of the data by the generic classifier.

24. The method of claim 23, further comprising:

establishing at least one confusion rule for the generic classifier, each confusion rule having at least a first portion and a second portion; and

building at least a two-way classifier for each of the confusion rules for utilization by at least one user-specific classifier, the two-way classifier comprising at least a first classification representing at least the first portion of the confusion rule and at least a second classification representing at least the second portion of the confusion rule.

25. The method of claim 24, the two-way classifier comprising, at least in part, at least one selected from the group consisting of an MLLR adapted density model, a direct density model, and a direct discriminative model.

26. The method of claim 24, the two-way classifier comprising models developed via a combination of likelihood from at least one adapted user-specific model and at least one generic model.

27. The method of claim 26, the combination comprising results from a multiplication process of likelihood from the adapted user-specific model and the generic model.

28. The method of claim 24, further comprising:

employing the generic classifier to determine when data is in the first classification;

utilizing the user-specific classifier to further refine data when the generic classifier determines data to be in the first classification; and

accepting the determination of the generic classifier when data is classified in a classification other than the first classification.

29. The method of claim 28, further comprising:

determining the number of characters in the second portion of the confusion rule when utilizing the user-specific classifier;

overruling the generic classifier when only one character is in the second portion of the confusion rule; and

classifying the character via the generic classifier among characters in the second portion of the confusion rule when more than one character is in the second portion of the confusion rule.

32. A device employing the method of claim 23 comprising at least one selected from the group consisting of a computer, a server, and a handheld electronic device.

34. A computer readable medium having stored thereon computer executable instructions for performing the method of claim 23.

30. A data packet transmitted between two or more computer components that facilitates handwriting recognition, the data packet comprising, at least in part, handwriting related information from a handwriting recognition system that utilizes a generic classifier and a user-specific classifier in conjunction to classify handwriting data.