Extrapolating false alarm rates in automatic speaker verification

TL;DR

This paper introduces generative models for extrapolating false alarm rates in automatic speaker verification, enabling performance prediction on large populations without additional data collection.

Contribution

It proposes a worst-case, generative modeling approach in the detection score space for reliable false alarm rate extrapolation in ASV systems.

Findings

Models enable sampling of new speakers for performance analysis.

Approach facilitates extrapolation without collecting new data.

Applicable to various ASV systems.

Abstract

Automatic speaker verification (ASV) vendors and corpus providers would both benefit from tools to reliably extrapolate performance metrics for large speaker populations without collecting new speakers. We address false alarm rate extrapolation under a worst-case model whereby an adversary identifies the closest impostor for a given target speaker from a large population. Our models are generative and allow sampling new speakers. The models are formulated in the ASV detection score space to facilitate analysis of arbitrary ASV systems.