# Adversarial Speaker Verification

## TL;DR

This paper introduces an adversarial training method for speaker verification that learns condition-invariant embeddings, improving robustness against environmental and SNR mismatches in speaker recognition tasks.

## Contribution

It proposes a novel adversarial multi-task training scheme to learn condition-invariant embeddings and extends it to multi-factorial scenarios for enhanced robustness.

## Key findings

- Achieves 8.8% relative EER reduction on known conditions
- Achieves 14.5% relative EER reduction on unknown conditions
- Demonstrates effectiveness on Microsoft Cortana speaker verification task

## Abstract

The use of deep networks to extract embeddings for speaker recognition has proven successfully. However, such embeddings are susceptible to performance degradation due to the mismatches among the training, enrollment, and test conditions. In this work, we propose an adversarial speaker verification (ASV) scheme to learn the condition-invariant deep embedding via adversarial multi-task training. In ASV, a speaker classification network and a condition identification network are jointly optimized to minimize the speaker classification loss and simultaneously mini-maximize the condition loss. The target labels of the condition network can be categorical (environment types) and continuous (SNR values). We further propose multi-factorial ASV to simultaneously suppress multiple factors that constitute the condition variability. Evaluated on a Microsoft Cortana text-dependent speaker verification task, the ASV achieves 8.8% and 14.5% relative improvements in equal error rates (EER) for known and unknown conditions, respectively.