Bayesian calibration for forensic evidence reporting
This addresses a specific calibration challenge in forensic speaker recognition, offering a principled improvement for practitioners in that domain.
The paper tackles the problem of forensic speaker recognition with limited background data for calibration by introducing a Bayesian solution that produces a Bayesian likelihood-ratio for evidence reporting. Experimental results on NIST SRE'12 scores show a clear advantage over traditional plugin calibration methods.
We introduce a Bayesian solution for the problem in forensic speaker recognition, where there may be very little background material for estimating score calibration parameters. We work within the Bayesian paradigm of evidence reporting and develop a principled probabilistic treatment of the problem, which results in a Bayesian likelihood-ratio as the vehicle for reporting weight of evidence. We show in contrast, that reporting a likelihood-ratio distribution does not solve this problem. Our solution is experimentally exercised on a simulated forensic scenario, using NIST SRE'12 scores, which demonstrates a clear advantage for the proposed method compared to the traditional plugin calibration recipe.