msSKA ‐ TDNNlite : Lightweight Speaker Verification with Multi‐Scale Selective Kernel and Channel‐Frequency Attention Article Swipe

PDF

Yuan Qiu , Rong Fei ·

YOU? · · 2025 · Open Access · · DOI: https://doi.org/10.1002/tee.70187

Deep neural networks have been widely used for speaker embedding extraction with remarkable results. In the speaker verification task, advanced network structures improve the performance of the model by introducing different attention mechanisms. However, it is still challenging to design a system that satisfies the lightweight and robustness requirements. In this paper, we propose a lightweight network based on channel‐frequency attention, namely msSKA‐TDNNlite. The msSKA‐C2D Block in the network simultaneously utilizes two attention mechanisms: Multi‐scale Selective Kernel Attention (msSKA) and C2D‐Att. The weights generated by these mechanisms can represent information in both the channel and frequency domains, with only a slight increase in parameters and computational cost. Finally, we conducted experiments on CN‐Celeb, a public benchmark dataset, and evaluated using test sets in both Chinese and English languages. The experimental results demonstrate that our method has fewer parameters than ECAPA‐TDNNLite and the low‐dimensional ECAPA‐TDNN, and achieves better performance on the Chinese and English test sets. © 2025 Institute of Electrical Engineers of Japan. Published by Wiley Periodicals LLC.

Related Topics

Truth And Reconciliation Commission Of Canada

Alanis Morissette

2025 Nba Draft

28 Years Later

Concepts

No concepts available.

Metadata

Type: article
Language: en
Landing Page: https://doi.org/10.1002/tee.70187
PDF: https://onlinelibrary.wiley.com/doi/pdfdirect/10.1002/tee.70187
OA Status: bronze
References: 11
OpenAlex ID: https://openalex.org/W4416338057

All OpenAlex metadata

Raw OpenAlex JSON

OpenAlex ID: https://openalex.org/W4416338057

Canonical identifier for this work in OpenAlex
DOI: https://doi.org/10.1002/tee.70187

Digital Object Identifier
Title: msSKA ‐ TDNNlite : Lightweight Speaker Verification with Multi‐Scale Selective Kernel and Channel‐Frequency Attention

Work title
Type: article

OpenAlex work type
Language: en

Primary language
Publication year: 2025

Year of publication
Publication date: 2025-11-18

Full publication date if available
Authors: Yuan Qiu, Rong Fei

List of authors in order
Landing page: https://doi.org/10.1002/tee.70187

Publisher landing page
PDF URL: https://onlinelibrary.wiley.com/doi/pdfdirect/10.1002/tee.70187

Direct link to full text PDF
Open access: Yes

Whether a free full text is available
OA status: bronze

Open access status per OpenAlex
OA URL: https://onlinelibrary.wiley.com/doi/pdfdirect/10.1002/tee.70187

Direct OA link when available
Cited by: 0

Total citation count in OpenAlex
References (count): 11

Number of works referenced by this work

Full payload

id	https://openalex.org/W4416338057
doi	https://doi.org/10.1002/tee.70187
ids.doi	https://doi.org/10.1002/tee.70187
ids.openalex	https://openalex.org/W4416338057
fwci
type	article
title	msSKA ‐ TDNNlite : Lightweight Speaker Verification with Multi‐Scale Selective Kernel and Channel‐Frequency Attention
biblio.issue
biblio.volume
biblio.last_page
biblio.first_page
is_xpac	False
apc_list.value	2900
apc_list.currency	USD
apc_list.value_usd	2900
apc_paid
language	en
locations[0].id	doi:10.1002/tee.70187
locations[0].is_oa	True
locations[0].source.id	https://openalex.org/S10731693
locations[0].source.issn	1931-4973, 1931-4981
locations[0].source.type	journal
locations[0].source.is_oa	False
locations[0].source.issn_l	1931-4973
locations[0].source.is_core	True
locations[0].source.is_in_doaj	False
locations[0].source.display_name	IEEJ Transactions on Electrical and Electronic Engineering
locations[0].source.host_organization	https://openalex.org/P4310320595
locations[0].source.host_organization_name	Wiley
locations[0].source.host_organization_lineage	https://openalex.org/P4310320595
locations[0].source.host_organization_lineage_names	Wiley
locations[0].license
locations[0].pdf_url	https://onlinelibrary.wiley.com/doi/pdfdirect/10.1002/tee.70187
locations[0].version	publishedVersion
locations[0].raw_type	journal-article
locations[0].license_id
locations[0].is_accepted	True
locations[0].is_published	True
locations[0].raw_source_name	IEEJ Transactions on Electrical and Electronic Engineering
locations[0].landing_page_url	https://doi.org/10.1002/tee.70187
indexed_in	crossref
authorships[0].author.id	https://openalex.org/A5101528828
authorships[0].author.orcid	https://orcid.org/0000-0002-4531-077X
authorships[0].author.display_name	Yuan Qiu
authorships[0].countries	CN
authorships[0].affiliations[0].institution_ids	https://openalex.org/I4210131919
authorships[0].affiliations[0].raw_affiliation_string	School of Computer Science and Engineering, Xi'an University of Technology, Xi'an, 710048 Shaanxi, China
authorships[0].institutions[0].id	https://openalex.org/I4210131919
authorships[0].institutions[0].ror	https://ror.org/038avdt50
authorships[0].institutions[0].type	education
authorships[0].institutions[0].lineage	https://openalex.org/I4210131919
authorships[0].institutions[0].country_code	CN
authorships[0].institutions[0].display_name	Xi'an University of Technology
authorships[0].author_position	first
authorships[0].raw_author_name	Yuan Qiu
authorships[0].is_corresponding	False
authorships[0].raw_affiliation_strings	School of Computer Science and Engineering, Xi'an University of Technology, Xi'an, 710048 Shaanxi, China
authorships[1].author.id	https://openalex.org/A5010752811
authorships[1].author.orcid	https://orcid.org/0000-0003-1892-8192
authorships[1].author.display_name	Rong Fei
authorships[1].countries	CN
authorships[1].affiliations[0].institution_ids	https://openalex.org/I4210131919
authorships[1].affiliations[0].raw_affiliation_string	School of Computer Science and Engineering, Xi'an University of Technology, Xi'an, 710048 Shaanxi, China
authorships[1].institutions[0].id	https://openalex.org/I4210131919
authorships[1].institutions[0].ror	https://ror.org/038avdt50
authorships[1].institutions[0].type	education
authorships[1].institutions[0].lineage	https://openalex.org/I4210131919
authorships[1].institutions[0].country_code	CN
authorships[1].institutions[0].display_name	Xi'an University of Technology
authorships[1].author_position	middle
authorships[1].raw_author_name	Rong Fei
authorships[1].is_corresponding	True
authorships[1].raw_affiliation_strings	School of Computer Science and Engineering, Xi'an University of Technology, Xi'an, 710048 Shaanxi, China
has_content.pdf	True
has_content.grobid_xml	False
is_paratext	False
open_access.is_oa	True
open_access.oa_url	https://onlinelibrary.wiley.com/doi/pdfdirect/10.1002/tee.70187
open_access.oa_status	bronze
open_access.any_repository_has_fulltext	False
created_date	2025-11-18T00:00:00
display_name	msSKA ‐ TDNNlite : Lightweight Speaker Verification with Multi‐Scale Selective Kernel and Channel‐Frequency Attention
has_fulltext	False
is_retracted	False
updated_date	2025-11-28T11:35:03.824165
primary_topic
cited_by_count	0
locations_count	1
best_oa_location.id	doi:10.1002/tee.70187
best_oa_location.is_oa	True
best_oa_location.source.id	https://openalex.org/S10731693
best_oa_location.source.issn	1931-4973, 1931-4981
best_oa_location.source.type	journal
best_oa_location.source.is_oa	False
best_oa_location.source.issn_l	1931-4973
best_oa_location.source.is_core	True
best_oa_location.source.is_in_doaj	False
best_oa_location.source.display_name	IEEJ Transactions on Electrical and Electronic Engineering
best_oa_location.source.host_organization	https://openalex.org/P4310320595
best_oa_location.source.host_organization_name	Wiley
best_oa_location.source.host_organization_lineage	https://openalex.org/P4310320595
best_oa_location.source.host_organization_lineage_names	Wiley
best_oa_location.license
best_oa_location.pdf_url	https://onlinelibrary.wiley.com/doi/pdfdirect/10.1002/tee.70187
best_oa_location.version	publishedVersion
best_oa_location.raw_type	journal-article
best_oa_location.license_id
best_oa_location.is_accepted	True
best_oa_location.is_published	True
best_oa_location.raw_source_name	IEEJ Transactions on Electrical and Electronic Engineering
best_oa_location.landing_page_url	https://doi.org/10.1002/tee.70187
primary_location.id	doi:10.1002/tee.70187
primary_location.is_oa	True
primary_location.source.id	https://openalex.org/S10731693
primary_location.source.issn	1931-4973, 1931-4981
primary_location.source.type	journal
primary_location.source.is_oa	False
primary_location.source.issn_l	1931-4973
primary_location.source.is_core	True
primary_location.source.is_in_doaj	False
primary_location.source.display_name	IEEJ Transactions on Electrical and Electronic Engineering
primary_location.source.host_organization	https://openalex.org/P4310320595
primary_location.source.host_organization_name	Wiley
primary_location.source.host_organization_lineage	https://openalex.org/P4310320595
primary_location.source.host_organization_lineage_names	Wiley
primary_location.license
primary_location.pdf_url	https://onlinelibrary.wiley.com/doi/pdfdirect/10.1002/tee.70187
primary_location.version	publishedVersion
primary_location.raw_type	journal-article
primary_location.license_id
primary_location.is_accepted	True
primary_location.is_published	True
primary_location.raw_source_name	IEEJ Transactions on Electrical and Electronic Engineering
primary_location.landing_page_url	https://doi.org/10.1002/tee.70187
publication_date	2025-11-18
publication_year	2025
referenced_works	https://openalex.org/W2103250316, https://openalex.org/W2760938034, https://openalex.org/W1589137271, https://openalex.org/W2150769028, https://openalex.org/W2890964092, https://openalex.org/W4319862221, https://openalex.org/W4206908380, https://openalex.org/W114517082, https://openalex.org/W3176969530, https://openalex.org/W2928165649, https://openalex.org/W2981087920
referenced_works_count	11
abstract_inverted_index.a	40, 54, 99, 113
abstract_inverted_index.In	14, 49
abstract_inverted_index.by	28, 84, 164
abstract_inverted_index.in	66, 90, 102, 122
abstract_inverted_index.is	35
abstract_inverted_index.it	34
abstract_inverted_index.of	25, 158, 161
abstract_inverted_index.on	58, 111, 148
abstract_inverted_index.to	38
abstract_inverted_index.we	52, 108
abstract_inverted_index.©	155
abstract_inverted_index.The	63, 81, 128
abstract_inverted_index.and	46, 79, 94, 104, 117, 125, 140, 144, 151
abstract_inverted_index.can	87
abstract_inverted_index.for	7
abstract_inverted_index.has	135
abstract_inverted_index.our	133
abstract_inverted_index.the	15, 23, 26, 44, 67, 92, 141, 149
abstract_inverted_index.two	71
abstract_inverted_index.2025	156
abstract_inverted_index.Deep	0
abstract_inverted_index.LLC.	167
abstract_inverted_index.been	4
abstract_inverted_index.both	91, 123
abstract_inverted_index.have	3
abstract_inverted_index.only	98
abstract_inverted_index.sets	121
abstract_inverted_index.test	120, 153
abstract_inverted_index.than	138
abstract_inverted_index.that	42, 132
abstract_inverted_index.this	50
abstract_inverted_index.used	6
abstract_inverted_index.with	11, 97
abstract_inverted_index.Block	65
abstract_inverted_index.Wiley	165
abstract_inverted_index.based	57
abstract_inverted_index.cost.	106
abstract_inverted_index.fewer	136
abstract_inverted_index.model	27
abstract_inverted_index.sets.	154
abstract_inverted_index.still	36
abstract_inverted_index.task,	18
abstract_inverted_index.these	85
abstract_inverted_index.using	119
abstract_inverted_index.Japan.	162
abstract_inverted_index.Kernel	76
abstract_inverted_index.better	146
abstract_inverted_index.design	39
abstract_inverted_index.method	134
abstract_inverted_index.namely	61
abstract_inverted_index.neural	1
abstract_inverted_index.paper,	51
abstract_inverted_index.public	114
abstract_inverted_index.slight	100
abstract_inverted_index.system	41
abstract_inverted_index.widely	5
abstract_inverted_index.(msSKA)	78
abstract_inverted_index.Chinese	124, 150
abstract_inverted_index.English	126, 152
abstract_inverted_index.channel	93
abstract_inverted_index.improve	22
abstract_inverted_index.network	20, 56, 68
abstract_inverted_index.propose	53
abstract_inverted_index.results	130
abstract_inverted_index.speaker	8, 16
abstract_inverted_index.weights	82
abstract_inverted_index.Finally,	107
abstract_inverted_index.However,	33
abstract_inverted_index.achieves	145
abstract_inverted_index.advanced	19
abstract_inverted_index.dataset,	116
abstract_inverted_index.domains,	96
abstract_inverted_index.increase	101
abstract_inverted_index.networks	2
abstract_inverted_index.results.	13
abstract_inverted_index.utilizes	70
abstract_inverted_index.Attention	77
abstract_inverted_index.Engineers	160
abstract_inverted_index.Institute	157
abstract_inverted_index.Published	163
abstract_inverted_index.Selective	75
abstract_inverted_index.attention	31, 72
abstract_inverted_index.benchmark	115
abstract_inverted_index.conducted	109
abstract_inverted_index.different	30
abstract_inverted_index.embedding	9
abstract_inverted_index.evaluated	118
abstract_inverted_index.frequency	95
abstract_inverted_index.generated	83
abstract_inverted_index.represent	88
abstract_inverted_index.satisfies	43
abstract_inverted_index.C2D‐Att.	80
abstract_inverted_index.Electrical	159
abstract_inverted_index.attention,	60
abstract_inverted_index.extraction	10
abstract_inverted_index.languages.	127
abstract_inverted_index.mechanisms	86
abstract_inverted_index.parameters	103, 137
abstract_inverted_index.remarkable	12
abstract_inverted_index.robustness	47
abstract_inverted_index.structures	21
abstract_inverted_index.CN‐Celeb,	112
abstract_inverted_index.Periodicals	166
abstract_inverted_index.challenging	37
abstract_inverted_index.demonstrate	131
abstract_inverted_index.experiments	110
abstract_inverted_index.information	89
abstract_inverted_index.introducing	29
abstract_inverted_index.lightweight	45, 55
abstract_inverted_index.mechanisms.	32
abstract_inverted_index.mechanisms:	73
abstract_inverted_index.msSKA‐C2D	64
abstract_inverted_index.performance	24, 147
abstract_inverted_index.experimental	129
abstract_inverted_index.verification	17
abstract_inverted_index.ECAPA‐TDNN,	143
abstract_inverted_index.Multi‐scale	74
abstract_inverted_index.computational	105
abstract_inverted_index.requirements.	48
abstract_inverted_index.simultaneously	69
abstract_inverted_index.ECAPA‐TDNNLite	139
abstract_inverted_index.low‐dimensional	142
abstract_inverted_index.msSKA‐TDNNlite.	62
abstract_inverted_index.channel‐frequency	59
cited_by_percentile_year
corresponding_author_ids	https://openalex.org/A5010752811
countries_distinct_count	1
institutions_distinct_count	2
corresponding_institution_ids	https://openalex.org/I4210131919
citation_normalized_percentile