Character recognition

• US 6,711,290 B2
• Filed: 01/19/2001
• Issued: 03/23/2004
• Est. Priority Date: 08/26/1998
• Status: Expired due to Term

• Current Assignee: Zi Decuma AB
• Original Assignee: Decuma AB (Microsoft Corporation)
• Inventors: Berthilsson, Rikard, Sparr, Gunnar
• Primary Examiner(s): CHANG, JON CARLTON
• Application Number:
  US09/764,341
• Publication Number:
  US 20010026639A1
• Time in Patent Office: 1,159 Days
• Field of Search: 382/119, 382/123, 382/187-189, 382/200-201, 382/216, 382/242, 382/203, 382/218, 382/217
• US Class Current: 382/186
• CPC Class Codes: G06V 30/1423 the instrument generating sequences of position coordinates corresponding to handwriting preprocessing or recognising digital ink G06V30/32

Abstract

A method of character recognition includes detecting a union of characters, preprocessing the union of characters, comparing the preprocessed union of characters with one or more template symbols, and applying a decision rule in order to either reject a template symbol or decide that the template symbol is included in the union of characters. Preprocessing the union of characters involves representing the union of characters as one or more curves, and parameterising said curve or curves, and regarding various classes of transformations forming one or more shapes for the curve or curves. The comparing operation involves forming one or more geometric proximity measures, and determining for every shape the values of said geometric proximity measures between said shape and correspondingly determined shapes for the template symbols. Finally, applying a decision rule involves selecting one or more template symbols in consideration of the values.

Claims

1. A method for character recognition comprising the steps of:

detecting a union of characters,

preprocessing the union of characters,

comparing the preprocessed union of characters with one or more template symbols, and

applying a decision rule in order to either reject a template symbol or decide that the template symbol is included in the union of characters, the step of preprocessing the union of characters comprising the steps of;

representing the union of characters as one or more curves, and

parameterising said curve or curves, characterised in that

the step of preprocessing the union of characters further comprises the step of forming, regarding various classes of transformations, one or more shapes for said curve or curves, wherein at least one of said classes of transformations is a subgroup of a group of all projective transformations, said subgroup comprising non-isometric transformations, and wherein said one or more shapes are invariant with regard to said classes of transformations, and

the step of comparing comprises the steps of;

forming one or more geometric proximity measures, wherein the step of forming one or more geometric proximity measures comprises selecting a scalar product that comprises derivatives of said curve or curves, and

determining for every shape the values of said geometric proximity measures between said shape and correspondingly determined shapes for the template symbols, and

the step of applying a decision rule comprises the step of;

selecting one or more template symbols in consideration of said values.

2. A method as claimed in claim 1, characterised in that the step of detecting a union of characters comprises the step of representing the union of characters as a set of points, and that the step of representing the union of characters as one or more curves comprises the steps of:

generating an ordered sequence of points from said set of points, and

interpolating between the points to generate said one or more curves.

3. A method as claimed in claim 2, characterized in that the step of parameterizing comprises:

arranging according to a rule of parameterization a function which follows the curve, and

sampling the function in a plurality of equidistant points.

4. A method as claimed in claim 3, characterised in that the rule of parameterisation is an arc length.

5. A method as claimed in claim 4, the union of characters being detected on a display on which it is written directly, characterised in that the step of detecting is carried out during the writing.

6. A method as claimed in claim 4, characterised in that the union of characters is detected in a data quantity that originates from a scanner.

7. A method as claimed in claim 6, characterised in that the step of preprocessing the union of characters comprises edge detecting the union of characters.

8. A method as claimed in claim 1, characterised in that the step of parameterising comprises the steps of:

arranging according to a rule of parameterisation a function which follows the curve, and sampling the function in a plurality of equidistant points.

9. A method as claimed in claim 8, characterised in that the rule of parameterisation is an arc length.

10. A method as claimed in claim 1, the union of characters being detected on a display on which it is written directly, characterised in that the step of detecting is carried out during the writing.

11. A method as claimed in claim 1, characterised in that the union of characters is detected in a data quantity that originates from a scanner.

12. A method as claimed in claim 11, characterised in that the step of preprocessing the union of characters comprises edge detecting the union of characters.

13. A method as claimed in claim 1, characterised in that the step of applying a decision rule comprises determining acceptance limits of the values of said proximity measures and selecting a template symbol only if at least one value related to the template symbol is within said acceptance limits.

14. A method as claimed in claim 13, characterised in that at least two template symbols have the same acceptance limits.

15. A method as claimed in claim 13, characterised in that the acceptance limits are determined on the basis of the fact that only similarity transformations are permitted.

16. A method as claimed in claim 1, characterised in that the step of applying a decision rule comprises carrying out a plausibility test of the selected template symbols.

17. The method of claim 1 wherein said steps are applied to a signature for verification or identification of said signature.

18. A method for character recognition comprising the steps of:

detecting a union of characters,

preprocessing the union of characters,

comparing the preprocessed union of characters with one or more template symbols, and

applying a decision rule in order to either reject a template symbol or decide that the template symbol is included in the union of characters, the step of preprocessing the union of characters comprising the steps of;

representing the union of characters as one or more curves, and parameterising said curve or curves, characterised in that

the step of preprocessing the union of characters further comprises the step of forming, regarding various classes of transformations, one or more shapes for said curve or curves, wherein at least one of said classes of transformations is a subgroup of a group of all projective transformations, said subgroup comprising non-isometric transformations, and wherein said one or more shapes are invariant with regard to said classes of transformations,

the step of comparing comprises the steps of;

forming one or more geometric proximity measures, and

determining for every shape the values of said geometric proximity measures between said shape and correspondingly determined shapes for the template symbols,

the step of applying a decision rule comprises the step of selecting one or more template symbols in consideration of said values, and

the step of applying a decision rule comprises determining acceptance limits of the values of said proximity measures and selecting a template symbol only if at least one value related to the template symbol is within said acceptance limits, and wherein individual acceptance limits are assigned to each template symbol.

19. A method for character recognition comprising the steps of:

detecting a union of characters,

preprocessing the union of characters,

comparing the preprocessed union of characters with one or more template symbols, and

applying a decision rule in order to either reject a template symbol or decide that the template symbol is included in the union of characters, the step of preprocessing the union of characters comprising the steps of;

representing the union of characters as one or more curves, and parameterising said curve or curves, characterised in that

the step of preprocessing the union of characters further comprises the step of forming, regarding various classes of transformations, one or more shapes for said curve or curves, wherein at least one of said classes of transformations is a subgroup of a group of all projective transformations, said subgroup comprising non-isometric transformations, and wherein said one or more shapes are invariant with regard to said classes of transformations,

the step of comparing comprises the steps of;

forming one or more geometric proximity measures, and

determining for every shape the values of said geometric proximity measures between said shape and correspondingly determined shapes for the template symbols,

the step of applying a decision rule comprises the step of selecting one or more template symbols in consideration of said values, and

the step of applying a decision rule comprises determining acceptance limits of the values of said proximity measures and selecting a template symbol only if at least one value related to the template symbol is within said acceptance limits, and that the method further comprises the step of reparameterising a parameterised curve if all the values of said proximity measures between the shape of the parameterised curve and the template symbols are beyond the acceptance limits so that one or more values of the corresponding proximity measure decreases between the template symbols and the shape of the reparameterised curve.

20. A method for character recognition comprising the steps of:

detecting a union of characters,

preprocessing the union of characters,

comparing the preprocessed union of characters with one or more template symbols, and

applying a decision rule in order to either reject a template symbol or decide that the template symbol is included in the union of characters, the step of preprocessing the union of characters comprising the steps of;

representing the union of characters as one or more curves, and parameterising said curve or curves, characterised in that

the step of preprocessing the union of characters further comprises the step of forming, regarding various classes of transformations, one or more shapes for said curve or curves, wherein at least one of said classes of transformations is a subgroup of a group of all projective transformations, said subgroup comprising non-isometric transformations, and wherein said one or more shapes are invariant with regard to said classes of transformations,

the step of comparing comprises the steps of;

forming one or more geometric proximity measures, and

determining for every shape the values of said geometric proximity measures between said shape and correspondingly determined shapes for the template symbols,

the step of applying a decision rule comprises the step of selecting one or more template symbols in consideration of said values,

the step of applying a decision rule comprises determining acceptance limits of the values of said proximity measures selecting a template symbol only if at least one value related to the template symbol is within said acceptance limits, and carrying out a plausibility test of the selected template symbols, and

the plausibility test is based on confidence sets.

21. A method for character recognition comprising the steps of:

detecting a union of characters,

preprocessing the union of characters,

comparing the preprocessed union of characters with one or more template symbols, and

applying a decision rule in order to either reject a template symbol or decide that the template symbol is included in the union of characters, the step of preprocessing the union of characters comprising the steps of;

representing the union of characters as one or more curves, and parameterising said curve or curves, characteriezed in that

the step of preprocessing the union of characters further comprises the step of forming, regarding various classes of transformations, one or more shapes for said curve or curves, wherein at least one of said classes of transformations is a subgroup of a group of all projective transformations, said subgroup comprising non-isometric transformations, and wherein said one or more shapes are invariant with regard to said classes of transformations, wherein the step of forming one or more shapes comprises forming, with regard to non-isometric transformations generated by compositions of translations, rotations, scalings, shears and reflections, a given shape for said curve or curves, such that said given shape is invariant with regard to all said non-isometric transformations, said given shape being given by

22. A method for character recognition comprising the steps of:

detecting a union of characters,

preprocessing the union of characters,

comparing the preprocessed union of characters with one or more template symbols, and

applying a decision rule in order to either reject a template symbol or decide that the template symbol is included in the union of characters, the step of preprocessing the union of characters comprising the steps of;

representing the union of characters as one or more curves, and parameterising said curve or curves, characterised in that

the step of preprocessing the union of characters further comprises the step of forming, regarding various classes of transformations, one or more shapes for said curve or curves, wherein at least one of said classes of transformations is a subgroup of a group of all projective transformations, said subgroup comprising non-isometric transformations, and wherein said one or more shapes are invariant with regard to said classes of transformations, wherein the step of forming one or more shapes comprises forming, with regard to affine transformations, a given shape for said curve or curves, such that said given shape is invariant with regard to all affine transformations said shape being given by

23. A method for character recognition comprising the steps of:

detecting a union of characters,

preprocessing the union of characters,

comparing the preprocessed union of characters with one or more template symbols, and

applying a decision rule in order to either reject a template symbol or decide that the template symbol is included in the union of characters, the step of preprocessing the union of characters comprising the steps of;

representing the union of characters as one or more curves, and parameterising said curve or curves, characterized in that

the step of preprocessing the union of characters further comprises the step of forming regarding various classes of transformations, one or more shapes for said curves or curves, wherein at least on of said classes of transformations is a subgroup of a group comprising non-isometric transformations, said subgroup comprising non-isometric transformations, and wherein said one or more shapes are invariant with regard to dais classes of transformations,

the step of comparing comprises the steps of;

forming one or more geometric proximity measures, and

determining for every shape the values of said geometric proximity measures between said shape and correspondingly determined shapes for the template symbols, and

the step of applying a decision rule comprises the step of selecting one or more template symbols in consideration of said values, and

wherein the step of forming one or more geometric proximity measures comprises selecting a scalar product that gives different weights to different parts of said curve or curves.