Publication Details

CA-MHFA: A Context-Aware Multi-Head Factorized Attentive Pooling for SSL-Based Speaker Verification

PENG Junyi, MOŠNER Ladislav, ZHANG Lin, PLCHOT Oldřich, STAFYLAKIS Themos, BURGET Lukáš and ČERNOCKÝ Jan. CA-MHFA: A Context-Aware Multi-Head Factorized Attentive Pooling for SSL-Based Speaker Verification. In: Proceedings of ICASSP 2025. Hyderabad: IEEE Biometric Council, 2025, pp. 1-5. ISBN 979-8-3503-6874-1. Available from: https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10889058
Czech title
CA-MHFA: Kontextově orientovaný extraktor informace o mluvčím pro ověřování mluvčího na základě samoučení
Type
conference paper
Language
english
Authors
Peng Junyi, Msc. Eng. (DCGM FIT BUT)
Mošner Ladislav, Ing. (DCGM FIT BUT)
Zhang Lin, Ph.D. (FIT BUT)
Plchot Oldřich, Ing., Ph.D. (DCGM FIT BUT)
Stafylakis Themos (OMILIA)
Burget Lukáš, doc. Ing., Ph.D. (DCGM FIT BUT)
Černocký Jan, prof. Dr. Ing. (DCGM FIT BUT)
URL
Keywords

Self-supervised learning, speaker verification, speaker extractor, pooling mechanism, speech classification

Abstract

Self-supervised learning (SSL) models for speaker verifica-
tion (SV) have gained significant attention in recent years. However,
existing SSL-based SV systems often struggle to capture local temporal
dependencies and generalize across different tasks. In this paper, we pro-
pose context-aware multi-head factorized attentive pooling (CA-MHFA),
a lightweight framework that incorporates contextual information from
surrounding frames. CA-MHFA leverages grouped, learnable queries to
effectively model contextual dependencies while maintaining efficiency
by sharing keys and values across groups. Experimental results on the
VoxCeleb dataset show that CA-MHFA achieves EERs of 0.42%, 0.48%,
and 0.96% on Vox1-O, Vox1-E, and Vox1-H, respectively, outperforming
complex models like WavLM-TDNN with fewer parameters and faster
convergence. Additionally, CA-MHFA demonstrates strong generalization
across multiple SSL models and tasks, including emotion recognition and
anti-spoofing, highlighting its robustness and versatility.

Published
2025
Pages
1-5
Proceedings
Proceedings of ICASSP 2025
Conference
ICASSP 2025, International Conference on Acoustics, Speech, and Signal Processing, Hyderabad, IN
ISBN
979-8-3503-6874-1
Publisher
IEEE Biometric Council
Place
Hyderabad, IN
DOI
10.1109/ICASSP49660.2025.10889058
BibTeX 
@INPROCEEDINGS{FITPUB13521,
   author = "Junyi Peng and Ladislav Mo\v{s}ner and Lin Zhang and Old\v{r}ich Plchot and Themos Stafylakis and Luk\'{a}\v{s} Burget and Jan \v{C}ernock\'{y}",
   title = "CA-MHFA: A Context-Aware Multi-Head Factorized Attentive Pooling for SSL-Based Speaker Verification",
   pages = "1--5",
   booktitle = "Proceedings of ICASSP 2025",
   year = 2025,
   location = "Hyderabad, IN",
   publisher = "IEEE Biometric Council",
   ISBN = "979-8-3503-6874-1",
   doi = "10.1109/ICASSP49660.2025.10889058",
   language = "english",
   url = "https://www.fit.vut.cz/research/publication/13521"
}
Back to top