PORTABLE DATA STORAGE MEDIUM FOR BIOMETRIC USER IDENTIFICATION

• US 20100054547A1
• Filed: 12/06/2007
• Published: 03/04/2010
• Est. Priority Date: 12/08/2006
• Status: Active Grant

• Current Assignee: Giesecke + Devrient Mobile Security GmBH (Giesecke & Devrient GmbH)
• Original Assignee: Giesecke & Devrient GmbH
• Inventors: Tagscherer, Michael
• Granted Patent: US 8,255,699 B2
• Time in Patent Office: Days
• US Class Current: 382/115
• CPC Class Codes: G06F 18/2137 based on criteria of topology preservation, e.g. multidimensional scaling or self-organising maps
• CPC Class Codes: G06F 18/28 Determining representative reference patterns, e.g. by averaging or distorting; Generating dictionaries
• CPC Class Codes: G06V 40/10 Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands

Abstract

A portable data carrier (1) comprises an executable training module (8) which provides in a memory (4, 5) of the data carrier (1) a reference model (10; M) for the biometric recognition of a user of the data carrier (1) by a recognition module (9), whereby a biometric comparison vector (16) which lies within the acceptance range (A) of the reference model (10; M) is accepted by the recognition module (9) as coming from the user. In so doing, the training module (8) defines the reference model (10; M) by model nodes (P, P0-P4) which are formed from biometric reference vectors (17; N) in each case coming from the user, whereby a new model node (P, P0-P4) is added to the reference model (10; M) when the associated reference vector (17; N) lies outside the acceptance range (A) of the reference model (10; M).

Claims

1-20. -20. (canceled)

21. A method for providing a reference model (10;

M) having an acceptance range (A) on a portable data carrier (1) for biometrically recognizing a user of the data carrier (1), wherein a biometric comparison vector (16) which lies within the acceptance range (A) of the reference model (10;

M) is accepted as coming from the user, comprising the steps of;

defining the reference model (10;

M) by model nodes (P, P0-P4) which are formed from biometric reference vectors (17;

N) in each case coming from the user, during an initial training phase, wherein for generating the reference model (10;

M) a multitude of biometric reference vectors (17;

N) is recorded, and wherein the reference model (10;

M) is defined piecewise by local models (L0-L3) each having a validity range (G0-G3), and adding a model node (P, P0-P4) to the reference model (10;

M), during an application phase of the reference model (10;

M) upon checking a comparison vector (16) when the comparison vector (16) lies outside the acceptance range (A) of the reference model (10;

M) and it is certain that the comparison vector (16) actually comes from the user.

22. The method according to claim 21, wherein a local model (L0-L3) is formed by at least one model node (P, P0-P4), and that for recognizing the user that local model (L0-L3) of the reference model (10;

M) is used in whose validity range (G0-G3) a biometric comparison vector (16) of the user lies.

23. The method according to claim 22, wherein each local model (L0-L3) is assigned in each case a tolerance range (T0-T3), the tolerance ranges (T0-T3) of all local models (L0-L3) forming the acceptance range (A) of the reference model (10;

M), and that a biometric comparison vector (16) which lies in the tolerance range (T0-T3) of a local model (L0-L3) is accepted as coming from the user.

24. The method according to claim 22, wherein each local model (L0-L3) is assigned in each case a validity range (G0-G3) which is determined by a probability distribution (W0-W2) which is assigned to the at least one model node (P, P0-P4) forming the local model (L0-L3), wherein the width of the probability distribution (W0-W2) of the model node (P, P0-P4) results from the probability distributions (W0-W2) of neighboring model nodes (P, P0-P4), and that the probability distribution (W0-W2) assigned to a model node (P, P0-P4) indicates a recognition quality (19) of a biometric comparison vector (16) accepted as coming from the user which lies within the validity range (G0-G3) of that local model (L0-L3) which is formed by at least the model node (P, P0-P4).

25. The method according to claim 24, wherein, upon addition of a model node (P, P0-P4) to the reference model (10;

M) the widths of the probability distribution (W0-W2) of the model node (P, P0-P4) and of the probability distributions (W0-W2) of neighboring model nodes (P, P0-P4) are adjusted to each other.

26. The method according to claim 22, wherein when the number of model nodes (P, P0-P4) forming a local model (L0-L3) exceeds a predetermined maximum number, a further model node (P, P0-P4) is not added to this local model (L0-L3) and/or the local model (L0-L3) is split into two new local models (L0-L3).

27. The method according to claim 22, wherein a decision about whether a biometric comparison vector (16) is a biometric reference vector (17;

N) coming from the user is effected by an authorized body (12) outside the data carrier (1).

35. The data carrier (1) according to claim 27, wherein

the recognition module (9) is configured to carry out a recognition of a biometric comparison vector (16) by matching the comparison vector (16) with a reference model (10;

M) of the user stored in the memory (4, 5); and

the training module (9) is configured to generate the reference model (10;

M) within the framework of an initial training phase of the recognition module (9) from a plurality of reference vectors (17;

N) received via the communication interface (6) and to store it in the memory (4, 5).

36. The data carrier (1) according to claim 27, wherein

the recognition module (9) is configured to carry out a recognition of a biometric comparison vector (16) by matching the comparison vector (16) with an adaptive reference model (10;

M) of the user stored in the memory (4, 5); and

the training module (8) is configured to continuously adjust the reference model (10;

M) within the framework of an application phase of the recognition module (9) with biometric reference vectors (17;

N) received via the communication interface (6) and to store it in the memory (4, 5).

38. The data carrier (1) according to claim 27, wherein the data carrier (1) is a chip card personalized for the user.

28. A portable data carrier (1) comprising a memory (4, 5), a communication interface (6), a processor (2), and a recognition module (9) and a training module (8) each executable by the processor (2), wherein the training module (8) is configured to provide in the memory (4, 5) a reference model (10;

M) having an acceptance range (A) for the biometric recognition of a user of the data carrier (1) by the recognition module (9), wherein a biometric comparison vector (16) which lies within the acceptance range (A) of the reference model (10;

M) is accepted by the recognition module (9) as coming from the user;

wherein the training module (8) is configured to define the reference model (10;

M) by model nodes (P, P0-P4) during an initial training phase,

wherein for generating the reference model (10;

M) a multitude of biometric reference vectors (17;

N) is recorded, which in each case are formed from biometric reference vectors (17;

N) received via the communication interface (6) and coming from the user,

wherein the training module (8) during an application phase of the reference model (10;

M) for checking a comparison vector (16) adds a model node (P, P0-P4) to the reference model (10;

M) when the comparison vector (16) lies outside the acceptance range (A) of the reference model (10;

M) and it is certain that the comparison vector (16) actually comes from the user.

29. The data carrier (1) according to claim 28, wherein

the training module (9) is configured to piecewise define the reference model (10;

M) by local models (L0-L3) each having a validity range (G0-G3), to form a local model (L0-L3) by at least one model node (P, P0-P4), and to store the local models (L0-L3) in the memory (4, 5); and

the recognition module (9) is configured to use that local model (L0-L3) of the reference model (10;

M) in whose validity range (G0-G3) a biometric comparison vector (16) of the user received via the communication interface (6) lies.

30. The data carrier (1) according to claim 29, wherein

the training module (8) is configured to assign a tolerance range (T0-T3) to each local model (L0-L3) and to form the acceptance range (A) of the reference model (10;

M) from the tolerance ranges (T0-T3) of all local models (L0-L3), and

the recognition module (9) is configured to accept as coming from the user a biometric comparison vector (16), received via the communication interface (9), which lies within the tolerance range (T0-T3) of a local model (L0-L3).

31. The data carrier (1) according to claim 29, wherein

the training module (9) is configured to assign to a local model (L0-L3) a validity range (G0-G3) which is determined by a probability distribution (W0-W2) which is assigned to the at least one model node (P, P0-P4) forming the local model (L0-L3), and to fix the width of the probability distribution (W0-W2) of the model node (P, P0-P4) in dependence on the probability distributions (W0-W2) of neighboring model nodes (P, P0-P4);

and

the recognition module (9) is configured to ascertain from the probability distribution (W0-W2) assigned to a model node (P, P0-P4) a recognition quality (19) of a biometric comparison vector (16) received via the communication interface (6) and accepted as coming from the user which lies within the validity range (G0-G3) of that local model (L0-L3) which is formed by at least the model node (P, P0-P4).

32. The data carrier (1) according to claim 31, wherein the training module (8) is set up such that upon addition of a model node (P, P0-P4) to the reference model (10;

M) it adjusts to each other the widths of the probability distribution (W0-W2) of the model node (P, P0-P4) and of the probability distributions (W0-W2) of neighboring model nodes (P, P0-P4).

33. The data carrier (1) according to claim 29, wherein the training module (8) is configured to not add a further model node (P, P0-P4) to the local model (L0-L3) and/or to split the local model (L0-L3) into two new local models (L0-L3), and to store them in the memory (4, 5), when the number of model nodes (P, P0-P4) forming a local model (L0-L3) exceeds a predetermined maximum number stored in the memory (4, 5).

34. The data carrier (1) according to claim 28, wherein the training module (8) is configured to store in the memory (4, 5), as a model node (P, P0-P4) to be added to the reference model (10;

M), the associated biometric reference vector (17;

N) of the user as a new model node (P, P0-P4).

37. The data carrier (1) according to claim 28, wherein the training module (8) is configured such that upon addition of a model node (P, P0-P4) to the reference model (10;

M) it gets information from the recognition module (9) that the associated reference vector (17;

N) lies outside the acceptance range (A) of the reference model (10;

M), and receives information from an authorized body (12) via the communication interface (6) that the associated reference vector (17;

N) comes from the user.