collaborators
Showing cs.SDShow all

6 papers · 1 filter

cs.SD2026

From Self-Supervised Speech Models to Mixture-of-Experts for Robust Anti-Spoofing

Hugo Daumain, Driss Matrouf, Khaled Khelif +1

Recent advances in speech generation have significantly improved the naturalness of synthetic speech, making spoofing detection increasingly challenging. A key limitation of curren…

cs.SD2026

Speaker-Invariant Representation Learning for Spoofing Detection via Gradient Reversal and A Variational Information Bottleneck

Anh-Tuan Dao, Driss Matrouf, Mickael Rouvier +1

Sophisticated generative speech technology can undermined the reliability of voice biometrics. While spoofing detection systems excel when assessed under in-domain conditions, gene…

cs.SD2026

A Comparison of SSL-Based Feature Extractors and Back-End Classifiers for Spoofing Detection: A Multi-Corpus Training and Cross-Linguistic Analysis

Anh-Tuan Dao, Driss Matrouf, Mickael Rouvier +1

Voice biometric systems face growing threats from spoofing attacks, yet the evaluation of detection models remains inconsistent across datasets. To investigate these unpredictable…

cs.SD2026

Assessing the Energy and Carbon Emissions of Neural Speaker Verification Model in Training and Inference

Hugo Leguillier, Driss Matrouf, Guillaume Lechien +1

Deep-learning speaker verification (SV) increasingly relies on deep neural network backbones, whose environmental impact remains largely undocumented. In this paper, we conduct an…

cs.SD2026

On Low-Bit Quantization Errors in Speaker Verification: Diagnostic and Mitigation

Hugo Leguillier, Driss Matrouf, Guillaume Lechien +1

Although low-bit quantization provides practical means to deploy speaker verification on resource-constrained devices, its effects on speaker verification performance remain poorly…

cs.SD2026

Assessing the Impact of Speaker Identity in Speech Spoofing Detection

Anh-Tuan Dao, Driss Matrouf, Nicholas Evans

Spoofing detection systems are typically trained using diverse recordings from multiple speakers, often assuming that the resulting embeddings are independent of speaker identity.…