Method for computer-adaptation of a reference data set on the basis of at least one input data set

• US 6,356,650 B1
• Filed: 08/06/1999
• Issued: 03/12/2002
• Est. Priority Date: 05/07/1997
• Status: Expired due to Term

• Current Assignee: Infineon Technologies AG
• Original Assignee: Siemens AG
• Inventors: Wirtz, Brigitte
• Primary Examiner(s): MEHTA, BHAVESH M
• Application Number:
  US09/367,022
• Time in Patent Office: 949 Days
• Field of Search: 382/119, 382/209, 382/217-218
• US Class Current: 382/119
• CPC Class Codes: G06F 18/28 Determining representative reference patterns, e.g. by averaging or distorting; Generating dictionaries
• CPC Class Codes: G06V 40/30 Writer recognition; Reading and verifying signatures
• CPC Class Codes: G07C 9/35 by means of a handwritten signature

Abstract

What the method assures is that natural changes in the signature of a person are not reflected in the form of increasing rejections within the framework of an authentification of this person on the basis of the person'"'"'s signature. “Natural” variation in the signature of the person is adapted to a reference signature, whereby an appropriate “distance” from a possible forged signature is adhered to so that adaptation is not undertaken in the direction of a potential forger. To that end, the signature is evaluated dependent on its distance from the reference signature and this evaluation criterion is divided into ranges that reflect the quality of the signature. Proceeding from the quality, an evaluation is possible as to whether or not an adaptation is implemented.

Claims

1. A method for computer-adaptation of a reference data set based on at least one input data set, comprising the steps of:

a) determining an evaluation criterion in view of a coincidence with the at least one reference data set for the input data set;

b) entering the input data set into a buffer memory and determining a new reference data set when the evaluation criterion lies within a predetermined, first range;

c) determining a next reference data set from the data sets from the buffer memory when the evaluation criterion lies within a predetermined, second range, and when a next reference data set is better than a previous reference data set, using the next reference data set and discarding the old reference data set, otherwise, when the next reference data set is poorer than the previous reference data set, discarding the next reference data set;

d) otherwise, not modifying the reference data set.

2. The method according to claim 1, wherein the reference data set is determined in that prescribable characteristic quantities of a prescribable plurality of data sets are averaged.

3. The method according to claim 1, wherein a data set is selected as reference data set from a prescribable plurality of data sets whose characteristic quantities best describe the selected data sets.

4. The method according to claim 1, wherein the buffer memory is a ring buffer having a prescribable plurality of data sets.

5. The method according to claim 1, wherein the input data set is an original data set when the evaluation criterion lies within the first range or second range or the input data set is a forgery when the evaluation criterion lies outside the first range or second range.

6. The method according to claim 1, wherein original data sets are distributed according to a probability distribution.

7. The method according to claim 6, wherein the probability distribution is a normal distribution with an anticipation value and a standard deviation.

8. The method according to claim 7, wherein the reference data set corresponds to the anticipation value of the original data sets.

9. The method according to claim 7, wherein the first range is determined by an interval

10. The method according to claim 9, wherein the input data set that lies in a fourth range between the second range and the third range with prescribable thresholds is recognized as original data set but the reference data set is not modified.

11. The method according to claim 1, wherein the data sets are signatures that are electronically registered and stored and that cover a prescribable set of characteristic quantities.