The IDLAB VoxSRC-20 Submission: Large Margin Fine-Tuning and Quality-Aware Score Calibration in DNN Based Speaker Verification

TL;DR

This paper introduces a large margin fine-tuning approach and a quality-aware score calibration method for DNN-based speaker verification, significantly improving robustness and calibration accuracy, leading to state-of-the-art results in VoxCeleb challenges.

Contribution

It presents a novel large margin fine-tuning strategy and a quality-aware score calibration method that enhance speaker verification performance and robustness.

Findings

Achieved state-of-the-art results on VoxCeleb1 test sets.

Enhanced system robustness with longer training utterances.

Contributed to winning VoxCeleb Speaker Recognition Challenge 2020.

Abstract

In this paper we propose and analyse a large margin fine-tuning strategy and a quality-aware score calibration in text-independent speaker verification. Large margin fine-tuning is a secondary training stage for DNN based speaker verification systems trained with margin-based loss functions. It enables the network to create more robust speaker embeddings by enabling the use of longer training utterances in combination with a more aggressive margin penalty. Score calibration is a common practice in speaker verification systems to map output scores to well-calibrated log-likelihood-ratios, which can be converted to interpretable probabilities. By including quality features in the calibration system, the decision thresholds of the evaluation metrics become quality-dependent and more consistent across varying trial conditions. Applying both enhancements on the ECAPA-TDNN architecture leads to state-of-the-art results on all publicly available VoxCeleb1 test sets and contributed to our winning submissions in the supervised verification tracks of the VoxCeleb Speaker Recognition Challenge 2020.