Biometrics
From Snehapadma SaaS for Biometric Attendance, Time Office and Payroll
Contents |
[edit] Definition
Biometrics is defined as science to automatically identify or verify identity of person based on physiological or behavioral characteristics.
[edit] Physiological characteristics
- Fingerprint
- Iris
- Hand Geometry
[edit] Behavioral characteristics
- Signature
- Face
- Voice
Though Face and Voice may seem physiological charactersitics, they depend heavily on behavior or mood of the person thus can be characterized as behavioral biometric technologies.
[edit] Key parameters of Biometrics
With so many biometric technologies available, it is important to understand and compare that makes these technologies effective. Some of them are:
- Uniqueness: That it should be unique to each person
- Permanence: That is should not change with time
- Intrusiveness: That is should be easy to collect and measure
- Accuracy and Speed: That it should be easy to process by system
- Economics: That it should be cost effective
[edit] Accuracy and Performance measurement
[edit] Enrollment Rate
This is defined as number of persons that are registered into the system as against intended to be enrolled into. E.g. if in case of a company having 100 employees, if 1 person could not enroll his fingerprint for some reason, the Enrollment Rate shall be 99%. This is mainly determined by hardware used for capturing fingerprint (or any biometrics for that matter). Better hardware shall have greater enrollment rate.
[edit] False Rejection Rate (FRR)
Biometrics is a percentage science. It provides that two fingerprint templates match each other to certain extend. If you want higher accuracy the algorithm scrutinizes fingerprint template in greater detail. This might leave even a legitimate fingerprint as unmatched because of various reason (e.g. uneven pressure is applied, or area of fingerprint scanned slightly differs from the area of registered fingerprint etc). The ratio of such legitimate biometrics not matching to total number of verifications done is called as False Rejection Rate or FRR. A good biometric technology should have lower FRR.
[edit] False Acceptance Rate (FAR)
When biometric is falsely accepted as legitimate one while it should not be, it is called as False Acceptance. The ratio of occurance of such false acceptance to total verifications performed is called as False Acceptance Rate or FAR.
[edit] Striking the Right Balance
Often for the same algorithm, if you reduce FAR, the FRR increases and vice a versa. These can also be said as measure of convenience and accuracy. Convenience meaning lower FRR and accuracy meaning lower FAR. Depending on application, you might want to go for convenience of accuracy. e.g. for high security area access control system, you might want to scrutinize a biometric more even when it means inconvenience to user in case he/she has to give multiple biometrics.
[edit] Fingerprint
Fingerprint is most commonly used and cost effective biometrics characteristic. A fingerprint reader scans fingerprint which has a pattern of ridge flow. This ridge flow some times ends abruptly or bifurcates to form to ridges. These are called minutiae. An algorithm makes note of such instance and the direction of ridge flow at the point. This is called as template. This stored in database (often in encrypted form).
At the time of verification, the fingerprint is scanned again to generate template as mentioned above. Then the template is matched against the stored template by considering relative positional and directional properties of minutiae. Algorithm arrives at score which determines how much the two fingerprints match. If this match is more than a certain threshold value, the templates are considered as match. If not, they are said not to match.
By increasing threshold value, you can increase the accuracy and reduce FAR at the cost of making it inconvenient or higher FAR. The reverse happens when you reduce the threshold value.
