DELULU: Discriminative Embedding Learning Using Latent Units for Speaker-Aware Self-Trained Speech Foundational Model

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Baali, Massa, Singh, Rita, Raj, Bhiksha
Natura: Preprint
Pubblicazione: 2025
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866917361341169664
author Baali, Massa
Singh, Rita
Raj, Bhiksha
author_facet Baali, Massa
Singh, Rita
Raj, Bhiksha
contents Self-supervised speech models have achieved remarkable success on content-driven tasks, yet they remain limited in capturing speaker-discriminative features critical for verification, diarization, and profiling applications. We introduce \textsc{DELULU}, a speaker-aware self-trained foundational model that addresses this limitation by incorporating speaker-informed structure into pseudo-label generation. DELULU leverages frame-level embeddings from ReDimNet, a state-of-the-art speaker verification model, to guide k-means clustering during pre-training, introducing a speaker-discriminative inductive bias that aligns representation learning with speaker identity. DELULU significantly outperforms prior SSL models across a range of speaker-centric tasks, achieving up to \textbf{62\% relative improvement} in equal error rate (EER) for speaker verification and consistent gains on zero-shot profiling tasks including gender, age, accent, and speaker counting; notably surpassing even its teacher model on zero-shot evaluations. Our findings demonstrate that \textbf{DELULU is a strong universal encoder for speaker-aware speech processing}, enabling superior performance without task-specific fine-tuning.
format Preprint
id arxiv_https___arxiv_org_abs_2510_17662
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle DELULU: Discriminative Embedding Learning Using Latent Units for Speaker-Aware Self-Trained Speech Foundational Model
Baali, Massa
Singh, Rita
Raj, Bhiksha
Sound
Computation and Language
Self-supervised speech models have achieved remarkable success on content-driven tasks, yet they remain limited in capturing speaker-discriminative features critical for verification, diarization, and profiling applications. We introduce \textsc{DELULU}, a speaker-aware self-trained foundational model that addresses this limitation by incorporating speaker-informed structure into pseudo-label generation. DELULU leverages frame-level embeddings from ReDimNet, a state-of-the-art speaker verification model, to guide k-means clustering during pre-training, introducing a speaker-discriminative inductive bias that aligns representation learning with speaker identity. DELULU significantly outperforms prior SSL models across a range of speaker-centric tasks, achieving up to \textbf{62\% relative improvement} in equal error rate (EER) for speaker verification and consistent gains on zero-shot profiling tasks including gender, age, accent, and speaker counting; notably surpassing even its teacher model on zero-shot evaluations. Our findings demonstrate that \textbf{DELULU is a strong universal encoder for speaker-aware speech processing}, enabling superior performance without task-specific fine-tuning.
title DELULU: Discriminative Embedding Learning Using Latent Units for Speaker-Aware Self-Trained Speech Foundational Model
topic Sound
Computation and Language
url https://arxiv.org/abs/2510.17662