Log In Sign Up

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

by   Jenthe Thienpondt, et al.

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.


page 1

page 2

page 3

page 4


The IDLAB VoxCeleb Speaker Recognition Challenge 2020 System Description

In this technical report we describe the IDLAB top-scoring submissions f...

Large Margin Softmax Loss for Speaker Verification

In neural network based speaker verification, speaker embedding is expec...

End-to-End Residual CNN with L-GM Loss Speaker Verification System

We propose an end-to-end speaker verification system based on the neural...

STC speaker recognition systems for the NIST SRE 2021

This paper presents a description of STC Ltd. systems submitted to the N...

Investigation of Different Calibration Methods for Deep Speaker Embedding based Verification Systems

Deep speaker embedding extractors have already become new state-of-the-a...

A Speaker Verification Backend with Robust Performance across Conditions

In this paper, we address the problem of speaker verification in conditi...

A discriminative condition-aware backend for speaker verification

We present a scoring approach for speaker verification that mimics the s...