Method and apparatus for choosing and evaluating sample size for biometric training process

• US 20080037832A1
• Filed: 08/10/2006
• Published: 02/14/2008
• Est. Priority Date: 08/10/2006
• Status: Active Grant

• Current Assignee: Louisiana Tech University Foundation (Louisiana Tech University)
• Original Assignee: Louisiana Tech University Foundation (Louisiana Tech University)
• Inventors: Phoha, Vir V.
• Granted Patent: US 7,809,170 B2
• Time in Patent Office: Days
• US Class Current: 382/115
• CPC Class Codes: G06F 18/25 Fusion techniques
• CPC Class Codes: G06V 40/70 Multimodal biometrics, e.g. combining information from different biometric modalities

Abstract

At least two biometric measurements of a person are collected, then a statistical measure based on the measurements is computed. The statistical measure is a bounded estimate of the discriminative power of a test based on the measurements. While the discriminative power is less than a target value, additional biometric measurements are collected. When enough measurements have been collected, a biometric template is constructed from the measurements and stored for use in future identifications. Systems and software to implement similar methods are also described and claimed.

Claims

1. A method comprising:

collecting a plurality of biometric measurements of a person;

computing a statistical measure based on the plurality of biometric measurements;

while the statistical measure is less than a target value, collecting a supplemental biometric measurement of the person and updating the statistical measure based on the supplemental biometric measurement; and

storing a biometric template constructed from the plurality of biometric measurements and any supplemental biometric measurements.

2. The method of claim 1, wherein

the statistical measure comprises an upper bound for an error variance.

3. The method of claim 1, wherein

the statistical measure is proportional to

$\frac{2}{}$

m 



σ

2 

Φ

2 Δ

2 ,

where

m is a number of elements in one biometric measurement,

σ

is an upper bound for an error variance,

Φ

₂ is a power of an F test, and

Δ

² is a maximum allowable difference.

4. The method of claim 1, further comprising:

computing an estimate of a discriminative power of a biometric identification system based on the plurality of biometric measurements and any supplemental biometric measurements; and

displaying the estimate of the discriminative power.

5. The method of claim 1, further comprising:

computing an estimate of a number of supplemental biometric measurements required to complete enrollment; and

displaying the estimate to the person.

6. The method of claim 1, further comprising:

collecting a candidate biometric measurement of a candidate;

comparing the candidate biometric measurement to the biometric template; and

granting access to a resource to the candidate if the candidate biometric measurement matches the biometric template.

7. The method of claim 1 wherein each biometric measurement comprises a plurality of keystroke timing measurements collected while the person types on a keyboard.

8. A computer-readable medium containing instructions to cause a programmable processor to perform operations comprising:

collecting a plurality of measurements of a characteristic of a user, each measurement including a plurality of scalar quantities;

adjusting the plurality of measurements to produce standard normal variables;

computing an estimate of a number of measurements required to achieve a target discriminative power of a test based on the standard normal variables; and

while a number of the plurality of measurements is less than the estimated number of measurements, collecting a supplemental measurement of the user and computing a newestimate of a number of measurements required to achieve the target discriminative power of the test based on the-standard normal variables and the supplemental measurement.

9. The computer-readable medium of claim 8, containing additional instructions to cause the programmable processor to perform operations comprising:

displaying the estimate to the user.

10. The computer-readable medium of claim 9 wherein displaying the estimate comprises displaying at least one of a number or a graphical representation of the estimate.

11. The computer-readable medium of claim 8 wherein the characteristic of the user is a typing rhythm of the user.

12. The computer-readable medium of claim 8 wherein the characteristic of the user is a measurement of a physical feature of the user.

13. The computer-readable medium of claim 12 wherein the physical feature is at least one of a hand geometry, a fingerprint, a retina image, or a voice recording.

14. A system comprising:

an input device to measure a characteristic of an enrolling user;

biometric sample collection logic to collect a plurality of measurements from the input device;

statistical analysis logic to compute a bounded estimate of a biometric identification capability based on the plurality of measurements; and

control logic to cause the biometric sample collection logic to collect a supplemental measurement until the bounded estimate reaches a target value.

15. The system of claim 14, further comprising:

a database to store a biometric template of the enrolling user prepared from the plurality of measurements and any supplemental measurements.

16. The system of claim 15, further comprising:

access control logic to compare a biometric sample of a prospective user with the biometric template and grant access to a resource if the comparison indicates that the prospective user is identical to the enrolling user.

17. The system of claim 14 wherein the characteristic is a typing rhythm.