Decision Making Based on Cohort Scores for Speaker Verification

TL;DR

This paper introduces a novel decision-making method for speaker verification that uses multiple cohort scores and a discriminative model, significantly improving performance over traditional single-score approaches.

Contribution

It proposes a discriminative decision model based on multiple cohort scores, enhancing speaker verification accuracy beyond conventional methods.

Findings

Deep neural network as decision maker yields substantial performance gains.

Using statistical features from cohort scores improves verification accuracy.

Method outperforms baseline single-score systems.

Abstract

Decision making is an important component in a speaker verification system. For the conventional GMM-UBM architecture, the decision is usually conducted based on the log likelihood ratio of the test utterance against the GMM of the claimed speaker and the UBM. This single-score decision is simple but tends to be sensitive to the complex variations in speech signals (e.g. text content, channel, speaking style, etc.). In this paper, we propose a decision making approach based on multiple scores derived from a set of cohort GMMs (cohort scores). Importantly, these cohort scores are not simply averaged as in conventional cohort methods; instead, we employ a powerful discriminative model as the decision maker. Experimental results show that the proposed method delivers substantial performance improvement over the baseline system, especially when a deep neural network (DNN) is used as the decision maker, and the DNN input involves some statistical features derived from the cohort scores.