Integrating feature extraction via local sequential embedding for automatic handwriting recognition

• US 8,977,059 B2
• Filed: 06/04/2012
• Issued: 03/10/2015
• Est. Priority Date: 06/03/2011
• Status: Active Grant

• Current Assignee: Apple Inc.
• Original Assignee: Apple Inc.
• Inventors: Bellegarda, Jerome R., Naik, Devang K., Dolfing, Jannes G. A.
• Primary Examiner(s): Le, Vu
• Assistant Examiner(s): WOLDEMARIAM, AKLILU K
• Application Number:
  US13/507,119
• Publication Number:
  US 20120308143A1
• Time in Patent Office: 1,009 Days
• Field of Search: 382/154, 382/187, 382/197, 382/202, 382/228, 382/188, 382/195, 382/301
• US Class Current: 382/197
• CPC Class Codes: G06V 30/347 Sampling; Contour coding; Stroke extraction

Abstract

Integrating features is disclosed, including: determining a value associated with a temporal feature for a point; determining a value associated with a spatial feature associated with the temporal feature; including the value associated with a spatial feature and the value associated with the temporal feature into a feature vector; and using the feature vector to decode for a character. Determining a transform is also disclosed, including: determining, for a point associated with a sequence of points, a set of points including: the point, a first subset of points of the sequence preceding a sequence position associated with the point, and a second subset of points following the sequence position associated with the point; and determining the transform associated with the point based at least in part on the set of points.

Claims

1. A system for integrating features, comprising:

a processor configured to;

determine a value associated with a temporal feature for a point associated with a sequence of points;

determine a value associated with a spatial feature for the point, computed from a set of points including;

the point, a first subset of points of the sequence preceding a sequence position associated with the point, and a second subset of points following the sequence position associated with the point;

include the value associated with the spatial feature and the value associated with the temporal feature in a single feature vector having both values; and

use at least the feature vector to decode for a character based at least in part on using one or more integrated recognition models configured to receive an input associated with at least a temporal feature and a spatial feature; and

a memory coupled to the processor and configured to provide the processor with instructions.

2. The system of claim 1, wherein the point comprises a sample point included in the sequence of points, and the sequence of points is derived from a handwriting input.

3. The system of claim 1, wherein the point comprises a pixel.

4. The system of claim 1, wherein the point is associated with a (x,y) coordinate.

5. The system of claim 1, wherein the value associated with the temporal feature is associated with at least a temporal or spatial neighborhood of points associated with the point.

6. The system of claim 1, wherein the one or more integrated recognition models are trained on an integration of spatial information and temporal information.

7. The system of claim 1, wherein the spatial feature may include one or more of the following:

chain or stroke code, sector occupancy, and pixel-level Rutovitz crossing number.

8. The system of claim 1, wherein the temporal feature may include one or more of the following:

position, velocity, and acceleration.

9. A method for integrating features, comprising:

at an electronic device having a processor and memory;

determining a value associated with a temporal feature for a point associated with a sequence of points;

determining a value associated with a spatial feature for the point, computed from a set of points including;

the point, a first subset of points of the sequence preceding a sequence position associated with the point, and a second subset of points following the sequence position associated with the point;

including the value associated with the spatial feature and the value associated with the temporal feature in a single feature vector having both values; and

using at least the feature vector to decode for a character based at least in part on using one or more integrated recognition models configured to receive an input associated with at least a temporal feature and a spatial feature.

10. The method of claim 9, wherein the point comprises a sample point included in the sequence of points, and the sequence of points is derived from a handwriting input.

11. The method of claim 9, wherein the value associated with the temporal feature is associated with at least a temporal or spatial neighborhood of points associated with the point.

12. The method of claim 9, wherein the one or more integrated recognition models are trained on an integration of spatial information and temporal information.

13. A computer program product for integrating features, the computer program product being embodied in a non-transitory computer readable medium and comprising computer instructions for:

determining a value associated with a temporal feature for a point associated with a sequence of points;

determining a value associated with a spatial feature for the point, computed from a set of points including;

the point, a first subset of points of the sequence preceding a sequence position associated with the point, and a second subset of points following the sequence position associated with the point;

including the value associated with the spatial feature and the value associated with the temporal feature in a single feature vector having both values; and

using at least the feature vector to decode for a character based at least in part on using one or more integrated recognition models configured to receive an input associated with at least a temporal feature and a spatial feature.

14. A system of determining a transform, comprising:

a processor configured to;

determine, for a point associated with a sequence of points, a set of points including;

the point, a first subset of points of the sequence preceding a sequence position associated with the point, and a second subset of points following the sequence position associated with the point;

determine the transform associated with the point based at least in part on the set of points;

include a value determined from the transform in a single feature vector including values corresponding to one or more of a spatial feature and a temporal feature associated with the point; and

use the feature vector to decode for a character using an integrated recognition model; and

a memory coupled to the processor and configured to provide the processor with instructions.

15. The system of claim 14, wherein the transform comprises a 1-dimensional discrete cosine transform.

16. The system of claim 14, wherein the transform comprises a Haar wavelets-based transform.

17. The system of claim 14, wherein the transform comprises a Fourier descriptor-based transform.

18. The system of claim 14, wherein the first subset of points comprises a same number of points as the second subset of points.

19. The system of claim 14, wherein the set of points is included within a sliding window associated with a size comprising a number of points present within the set of points.

20. The system of claim 14, wherein the point comprises a pixel.

21. A method of determining a transform, comprising:

at an electronic device having a processor and memory;

determining, for a point associated with a sequence of points, a set of points including;

the point, a first subset of points of the sequence preceding a sequence position associated with the point, and a second subset of points following the sequence position associated with the point;

determining the transform associated with the point based at least in part on the set of points;

including a value determined from the transform in a single feature vector including values corresponding to one or more of a spatial feature and a temporal feature associated with the point; and

using the feature vector to decode for a character using an integrated recognition model.

22. The method of claim 21, wherein the first subset of points comprises a same number of points as the second subset of points.

23. The method of claim 21, wherein the set of points is included within a sliding window associated with a size comprising a number of points present within the set of points.

24. A computer program product for determining a transform, the computer program product being embodied in a non-transitory computer readable medium and comprising computer instructions for:

determining, for a point associated with a sequence of points, a set of points including;

the point, a first subset of points of the sequence preceding a sequence position associated with the point, and a second subset of points following the sequence position associated with the point;

determining the transform associated with the point based at least in part on the set of points;

including a value determined from the transform in a single feature vector including values corresponding to one or more of a spatial feature and a temporal feature associated with the point; and

using the feature vector to decode for a character using an integrated recognition model.