how can I calculate the False Rejection Rate (FRR) and the False Acceptance Rate (FAR) for biometrics applications?

Hello

In biometrics applications, we must measure the False Rejection Rate (FRR) and the False Acceptance Rate (FAR).

I'm confused about this

If I use a database containing 100 people, each person has 8 images, 6 images used as a templates in the database, and two images used for testing purposes.

So I have 600 images as templates in the database and 200 images used for testing.

For the False Rejection Rate (FRR), do I divide the number of original people who have been rejected to 200?

For the False Acceptance Rate (FAR), do I divide the number of fraudsters who have been accepted to 800 or 600?

I'm a little confused.

I need some one to clarify this issue.

Thank you so much.

Answer 1

[KALYAN ACHARJYA]

If you have test 100 images to test the biometric security system.

Suppose if it wrongly accepts 10 users to access (unauthorized person), the FAR is 10% Those are wrong accepted (Those are unauthorized)

On the other hand, in FRR

The number of users wrongly decline the access, but actually, they are authorized to access

Moder Detail read here

[wisam kh]

thank you sir

There are those who say that the false acceptance rate, is derived by dividing the number of times the matching with the images in the database. In my case, each image is matched with 600 images in the database. Is this true or do I divide by 200 which is the number of test samples?

[KALYAN ACHARJYA]

FAR=Number of false acceptance/Nos of identification attempts (Testing Data);

FRR=Number of false rejection/Nos of identification attempts (Testing data);

Hope your doubt is cleared now.

[wisam kh]

Thank you so much.

Now things are clear.

[KALYAN ACHARJYA]

Welcome @Wisam

Happy to Help.