Whisper Speaker Identification: Leveraging Pre-Trained Multilingual Transformers for Robust Speaker Embeddings

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Emon, Jakaria Islam, Salek, Md Abu, Alam, Kazi Tamanna
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866917956029513728
author Emon, Jakaria Islam
Salek, Md Abu
Alam, Kazi Tamanna
author_facet Emon, Jakaria Islam
Salek, Md Abu
Alam, Kazi Tamanna
contents Speaker identification in multilingual settings presents unique challenges, particularly when conventional models are predominantly trained on English data. In this paper, we propose WSI (Whisper Speaker Identification), a framework that repurposes the encoder of the Whisper automatic speech recognition model pre trained on extensive multilingual data to generate robust speaker embeddings via a joint loss optimization strategy that leverages online hard triplet mining and self supervised Normalized Temperature-scaled Cross Entropy loss. By capitalizing on Whisper language-agnostic acoustic representations, our approach effectively distinguishes speakers across diverse languages and recording conditions. Extensive evaluations on multiple corpora, including VoxTube (multilingual), JVS (Japanese), CallHome (German, Spanish, Chinese, and Japanese), and Voxconverse (English), demonstrate that WSI consistently outperforms state-of-the-art baselines, namely Pyannote Embedding, ECAPA TDNN, and Xvector, in terms of lower equal error rates and higher AUC scores. These results validate our hypothesis that a multilingual pre-trained ASR encoder, combined with joint loss optimization, substantially improves speaker identification performance in non-English languages.
format Preprint
id arxiv_https___arxiv_org_abs_2503_10446
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Whisper Speaker Identification: Leveraging Pre-Trained Multilingual Transformers for Robust Speaker Embeddings
Emon, Jakaria Islam
Salek, Md Abu
Alam, Kazi Tamanna
Sound
Artificial Intelligence
Audio and Speech Processing
I.2
Speaker identification in multilingual settings presents unique challenges, particularly when conventional models are predominantly trained on English data. In this paper, we propose WSI (Whisper Speaker Identification), a framework that repurposes the encoder of the Whisper automatic speech recognition model pre trained on extensive multilingual data to generate robust speaker embeddings via a joint loss optimization strategy that leverages online hard triplet mining and self supervised Normalized Temperature-scaled Cross Entropy loss. By capitalizing on Whisper language-agnostic acoustic representations, our approach effectively distinguishes speakers across diverse languages and recording conditions. Extensive evaluations on multiple corpora, including VoxTube (multilingual), JVS (Japanese), CallHome (German, Spanish, Chinese, and Japanese), and Voxconverse (English), demonstrate that WSI consistently outperforms state-of-the-art baselines, namely Pyannote Embedding, ECAPA TDNN, and Xvector, in terms of lower equal error rates and higher AUC scores. These results validate our hypothesis that a multilingual pre-trained ASR encoder, combined with joint loss optimization, substantially improves speaker identification performance in non-English languages.
title Whisper Speaker Identification: Leveraging Pre-Trained Multilingual Transformers for Robust Speaker Embeddings
topic Sound
Artificial Intelligence
Audio and Speech Processing
I.2
url https://arxiv.org/abs/2503.10446