Knol will be unavailable during scheduled maintenance starting at Mon, 09 Nov 2009 18:30:00 GMT. We expect the maintenance to be completed at Mon, 09 Nov 2009 20:00:00 GMT.
Version: Baidi441

Biometrics

Identification Techniques for System

Humans identify people they see using variety of clues that brains process fast. Systems require specific inputs that is analyzed in specific manner. In this knol, we will be discussing these inputs, methods of acquiring and processing them.


Introduction


Biometrics technology is used to let computers identify people so that they are allowed to certain restricted area or application, pay bills, or just to let system make note of their presence. It does not require you to remember passwords, or carry cards that can be forgotten, lost or stolen.

Crudely put, biometrics refers to measurements of biological features.  In the systems context it is defined as automated methods of recognizing humans based on physiological or behavioral characteristics uniquely.

Automated


The biometrics system has to be automated. It should not have any manual intervention in acquiring input, or processing it. This means there must be a sensor to grab the characteristic, and processor to process the grabbed characteristic. In large scale systems such as AFIS used by most of the law enforcement agencies, the automated identification is used to shortlist the candidate matches with input. The final decision is taken by human agent.

Recognition


Recognition (re + cognition) implies that biometric system must have stored data about the characteristics. This data is matched systematically with the sensed data using some algorithm. These algorithms are specific to sensing technology and characteristic used for recognition. There are two types of recognition process. Identification and Verification.

Identification


Matching is done between the sensed data and stored data. If sensed data is matched with all (or multiple) records of stored data, it is called as Identification. Often it is also referred as 1:N matching process

Verification


In verification process, the stored data is filtered so that algorithm tries to match sensed data with only one record of stored data. This is referred as 1:1 matching process.

Physiological Characteristics


The third term of significance is physiological characteristics. These are the characteristics that are present by virtue of being part of your body. This do not change with your behavior or mood but might change with the age. examples are iris, fingerprint, face, gait, palm, ...

Behavioral Characteristics


These characteristics depend on your behavior. E.g. voice biometrics, signature, keystroke  recognition falls in this category. Basic difference with physiological characteristics is that you have choice to change these biometrics are per your wish.

Uniqueness


Of course the purpose of this is to recognize the person uniquely. To put forward the point, no of fingers is a physiological characteristic that can be sensed automatically and processed to matched. Since this characteristics will identify with multiple  people (in fact all with exceptions like Hritik Roshan) and hence can not be classified as biometrics. As against this, all the characteristics that are mentioned above can be fairly found to be unique to the person.

Choice of Biometrics


So, now I am planning to use biometrics for my application, what should I choose? Unfortunately, the asnwers, as of now, is not simple. You need to factor in various considerations before like price, maturity of the technology. Some others are:

Convenience or Security


What is prime purpose of using biometrics? Is it convinience, so that people do not have to remember the password or carry the card? Or is it that you want to put it for highly sensitive restricted area? Accordingly you can play with terms called False Acceptance Rate or FAR and False Rejection Rate called FRR.

FAR


False acceptance rate is chance that wrong person will be identified after offering credentials to the system. It is advised that you should put up system with low FAR to restrict entry to highly sensitive restricted area.

FRR


False Rejection Rate is change that even correc person will not be identified by the system. When convinience is priority, it can be anoying if person has to give biometrics again and again because of high FRR. In such cases, biometrics which has lower FRR should be used.

The Trade Off


For any biometric system, as your try and tweak algorithm to reduce FAR, the FRR goes up and vice versa. In such case, you should see do you want convininenc or security, and then accordingly tweak the algorithm.

Overt or Covert


Imagine a biometric system which is scannign face of each person entering into the mall or station and then running background search against wanted database. In such a case, the person to be identified does not know whether his characteristics are being recorded or not. In such a case, non intrusive biometrics like face or gait can be used. Opposite case is when person is availing ration service where he or she can get feedback and improve sensing of biometrics. In such a case, fingerprint or iris can be used. It is particulrly worth noting that behavioral biometrics be used in overt mechanism while non-intrusive physiological biometrics be used in covert mechanism.

Comments

Harshal Baviskar
Harshal Baviskar
Entrepreneur at Snehapadma Software Technologies Pvt Ltd
Pune
Article rating:
Your rating:
All Rights Reserved.
Version: 6
Versions
Last edited: Jul 24, 2008 11:56 AM.

Activity for this knol

This week:

21pageviews

Totals:

311pageviews