Character recognition

• US 7,139,430 B2
• Filed: 03/09/2004
• Issued: 11/21/2006
• Est. Priority Date: 08/26/1998
• Status: Expired due to Term

• Current Assignee: Cerence Operating Company (Cerence Inc.)
• Original Assignee: Zi Decuma AB
• Inventors: Sparr, Gunnar, Berthilsson, Rikard
• Primary Examiner(s): Bali, Vikkram
• Application Number:
  US10/795,245
• Publication Number:
  US 20040234129A1
• Time in Patent Office: 987 Days
• Field of Search: 382/119, 382/123, 382186-189, 382200-203, 382216-218, 382/242
• 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

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 to either reject a template symbol or decide that the template symbol is included in the union of characters. Such preprocessing involves representing the union of characters as one or more curves, and parameterizing the curve(s); and, regarding various classes of transformation, forming one or more shapes for the curve(s). The comparing operation involves forming one or more geometric proximity measures, and determining for every shape the values of those measures between the shape and correspondingly determined shapes for the template symbols. 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

parameterizing 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 curve or curves, wherein said one or more shapes for said curve or curves are defined as

d(φ

)={ψ

ψ

=g(φ

),g ε

G}

where φ

is said parameterized curve, d(φ

) is said shape for said curve, and G is said class of transformations,

said shape being invariant with respect to said class of transformations, and that the step of comparing comprises the steps of;

forming one or more geometric proximity measures,

determining for every shape the values of said geometric proximity measures between said shape and correspondingly determined shapes for the template symbols, and that the step of applying a decision rule comprises the step of;

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

3. A method as claimed in claim 1, characterized in that the step of parameterizing comprises the steps of:

arranging according to a convenient 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, characterized in that the rule of parameterization is an arc length.

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

6. A method as claimed in claim 1, characterized 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, characterized 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, characterized 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.

9. A method as claimed in claim 8, characterized in that individual;

acceptance limits are assigned to each template symbol.

10. A method as claimed in claim 8, characterized in that at least two template symbols have the same acceptance limits.

11. A method as claimed in claim 8, characterized in that the step of reparameterizing a parameterized curve is all the values of said proximity measures between the shape of the parameterized 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 reparameterized curve.

12. A method as claimed in claim 8, characterized in that the acceptance limits are determined on the basis of the fact that only similarity transformations are permitted.

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

14. A method as claimed in claim 13, characterized in that the plausibility test is based on the confidence sets.

15. A method as claimed in claim 1, wherein the step of forming one or more shapes comprises forming, with regard to positive similarity transformations, a given shape for said curve or curves, such that said given shape is invariant with regard to all said positive similarity transformations.

16. A method as claimed in claim 15, wherein said given shape is given by

s(φ

)=linhull ({(φ

₁, φ

₂), (−

φ

₂, φ

₁), (1,

     0), (0,

     1)}

where φ

is said parameterized curve or curves, and φ

is parameterized in two parts φ

₁ and φ

₂ representing two coordinates of points on said curve or curves.

17. A method as claimed in claim 1, 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.

18. A method as claimed in claim 17, wherein said given shape is given by

s(φ

)=linhull (φ

₁, φ

₂,

     1)

where φ

is said parameterized curve or curves, and φ

is parameterized in two parts φ

₁ and φ

₂ representing two coordinates or points on said curve or curves.

19. A method as claimed in claim 1, wherein the step of forming one or more geometric proximity measures comprises selecting a scalar product that comprises derivatives of said curve or curves.

20. A method as claims in claim 1, 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.

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

2. A method as claimed in 1, characterized 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.