Multi-Task Transformer for Explainable Speech Deepfake Detection via Formant Modeling
Sound
2026-01-23 v2
Abstract
In this work, we introduce a multi-task transformer for speech deepfake detection, capable of predicting formant trajectories and voicing patterns over time, ultimately classifying speech as real or fake, and highlighting whether its decisions rely more on voiced or unvoiced regions. Building on a prior speaker-formant transformer architecture, we streamline the model with an improved input segmentation strategy, redesign the decoding process, and integrate built-in explainability. Compared to the baseline, our model requires fewer parameters, trains faster, and provides better interpretability, without sacrificing prediction performance.
Cite
@article{arxiv.2601.14850,
title = {Multi-Task Transformer for Explainable Speech Deepfake Detection via Formant Modeling},
author = {Viola Negroni and Luca Cuccovillo and Paolo Bestagini and Patrick Aichroth and Stefano Tubaro},
journal= {arXiv preprint arXiv:2601.14850},
year = {2026}
}
Comments
Accepted @ IEEE ICASSP 2026