When enough is enough: early stopping of biometrics error rate testing

Testing of biometric devices to determine whether or not their error rates meet a specified threshold is often an important criterion for determining whether or not a device will be deployed. One consideration for this is how much testing is to be completed. In this paper we apply sequential testing methodology to testing whether or not a biometric device has an error rate below a certain bound. This approach has the advantage that the testing can be stopped at any time. A decision can be made once there is enough evidence to support or reject the hypothesis of interest. Application is made to biometric error rate data for three different modalities -facial recognition, hand geometry and fingerprints. Our results show that this approach can produce more efficient and less costly tests of biometric devices.