Method and mechanism to reduce handwriting recognizer errors using multiple decision trees

• US 7,379,597 B2
• Filed: 12/05/2005
• Issued: 05/27/2008
• Est. Priority Date: 06/06/1997
• Status: Expired due to Fees

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Hullender, Gregory N., Haluptzok, Patrick M., Bennett, John R.
• Primary Examiner(s): Carter; Aaron W
• Application Number:
  US11/294,260
• Publication Number:
  US 20060072825A1
• Time in Patent Office: 904 Days
• Field of Search: 382185-190, 382226-227, 382/157, 345/176
• US Class Current: 382/187
• CPC Class Codes: G06F 18/254 of classification results, e.g. of results related to same input data
• CPC Class Codes: G06V 30/347 Sampling; Contour coding; Stroke extraction
• CPC Class Codes: G06V 30/373 using a special pattern or subpattern alphabet

Abstract

An improved method and mechanism for recognizing chirographs (handwritten characters) input into a computer system. A primary recognizer is provided for converting chirographs to code points, and secondary recognizers such as binary CART trees are developed and trained to differentiate chirographs which produce certain code points at the primary recognizer. Each such secondary recognizer is associated with each selected code point. When a chirograph is received, the chirograph is provided to the primary recognizer whereby a code point corresponding thereto is received. If the code point corresponds to one of the secondary recognizers, the chirograph is passed to the secondary recognizer, and a code point is returned from the secondary recognizer. If not, the code point provided by the primary recognizer is returned. The invention sets forth an automated process for training the CART trees and for optimizing the recognition mechanism by discarding CART trees which do not improve on the recognition accuracy of the primary recognizer.

Claims

1. A computer-readable medium having computer-executable instructions, which when executed perform steps, comprising:

receiving a chirograph;

providing the chirograph to a primary recognizer and receiving a shape index therefrom;

determining whether one of a plurality of secondary recognizers is associated with the shape index, and if so, selecting that secondary recognizer as a selected secondary recognizer, wherein the plurality of secondary recognizers comprises CART trees that have been trained by providing a first training set comprising a plurality of chirographs and actual code points for each chirograph, applying a plurality of questions to the chirograph, determining a best set of questions in a best order, removing questions determined to be ineffective, and removing CART trees that do not improve the recognition of the characters compared to the primary recognizer; and

passing the chirograph to the selected secondary recognizer and returning a code point from the secondary recognizer.

2. The computer-readable medium of claim 1 wherein the shape index comprises a code point.

14. The computer-readable medium of claim 2 having computer-executable instructions, which when executed perform steps, further comprising:

receiving at least one alternative code points from the primary recognizer;

determining whether one of the plurality of secondary recognizers is associated with at least one of the alternative code points, and if so, selecting each secondary recognizer associated with an alternative code point as an alternative selected secondary recognizer for the respective alternative code point, wherein the plurality of secondary recognizers comprises CART trees that have been trained by providing a first training set comprising a plurality of chirographs and actual code points for each chirograph, applying a plurality of questions to the chirograph, determining a best set of questions in a best order, removing questions determined to be ineffective, and removing CART trees that do not improve the recognition of the characters compared to the primary recognizer;

passing the chirograph to each secondary recognizer associated with the respective alternative code and returning code points from the alternative secondary recognizers; and

sorting the returned code points according to the probability of a correct match.

3. A computer-readable medium having computer-executable instructions, which when executed perform steps, comprising:

providing a received chirograph to a primary recognizer to make a first decision as to a shape index that corresponds to the received chirograph;

selecting a secondary recognizer from a plurality of available secondary recognizers based on the shape index, wherein the plurality of secondary recognizers comprise CART trees that have been trained by providing a first training set comprising a plurality of chirographs and actual code points for each chirograph, applying a plurality of questions to the chirograph, determining a best set of questions in a best order, removing questions determined to be ineffective, and removing CART trees that do not improve the recognition of the characters compared to the primary recognizer;

providing the chirograph to the secondary recognizer, the secondary recognizer determining a recognition result independent of the shape index provided by the primary recognizer; and

returning the recognition result from the secondary recognizer.

4. The computer-readable medium of claim 3 wherein the shape index comprises a code point.

5. The computer-readable medium of claim 3 wherein the recognition result comprises a code point.

6. The computer-readable medium of claim 3 wherein the shape index does not correspond to a code point.

7. The computer-readable medium of claim 3 wherein each shape index that the primary recognizer is capable of outputting has a unique secondary recognizer associated therewith.

8. In a computing environment having at least one computing device, a system, comprising:

a primary recognizer that determines a shape index from a chirograph;

a plurality of secondary recognizers, each secondary recognizer corresponding to a shape index, wherein the plurality of secondary recognizers comprise CART trees that have been trained by providing a first training set comprising a plurality of chirographs and actual code points for each chirograph, applying a plurality of questions to the chirograph, determining a best set of questions in a best order, removing questions determined to be ineffective, and removing CART trees that do not improve the recognition of the characters compared to the primary recognizer;

an interface that receives a chirograph and provide it to the primary recognizer, the primary recognizer causing selection of a selected secondary recognizer based on a determined shape index corresponding to the chirograph, and

the selected secondary recognizer including means for determining a recognition result from the chirograph and returning the recognition result, wherein the returned recognition result need not correspond to a value of the shape index determined by the primary recognizer.

9. The system of claim 8 wherein the shape index comprises a single code point.

10. The system of claim 8 wherein the shape index comprises a single code point that differs from the returned code point.

11. The system of claim 8 wherein the recognition result comprises a single code point.

12. The system of claim 8 wherein the shape index does not correspond to a code point.

13. The system of claim 8 wherein the recognition information received from the primary recognizer does not correspond to a code point, and wherein the recognition result comprises a single code point.

15. The system of claim 8, wherein the interface that receives a chirograph comprises a touch pad device.