Squeezing value of cross-domain labels: a decoupled scoring approach for speaker verification

TL;DR

This paper investigates domain mismatch issues in speaker verification, revealing that simply adding cross-domain data is ineffective, and proposes a decoupled scoring method that optimally utilizes cross-domain labels for improved performance.

Contribution

The paper introduces a novel decoupled scoring approach that effectively leverages cross-domain labels to address domain mismatch in speaker verification systems.

Findings

Adding cross-domain data alone does not improve performance under mismatch.

The proposed method maximizes the utility of cross-domain labels for verification.

Experimental results demonstrate the approach's effectiveness in cross-channel tests.

Abstract

Domain mismatch often occurs in real applications and causes serious performance reduction on speaker verification systems. The common wisdom is to collect cross-domain data and train a multi-domain PLDA model, with the hope to learn a domain-independent speaker subspace. In this paper, we firstly present an empirical study to show that simply adding cross-domain data does not help performance in conditions with enrollment-test mismatch. Careful analysis shows that this striking result is caused by the incoherent statistics between the enrollment and test conditions. Based on this analysis, we present a decoupled scoring approach that can maximally squeeze the value of cross-domain labels and obtain optimal verification scores when the enrollment and test are mismatched. When the statistics are coherent, the new formulation falls back to the conventional PLDA. Experimental results on cross-channel test show that the proposed approach is highly effective and is a principle solution to domain mismatch.