CNN-based Text-independent Automatic Speaker Identification Using Short Utterances

Mandana Fasounaki; Emirhan Burak Yüce; Serkan Öncül; Gökhan Ince

doi:10.1109/UBMK52708.2021.9559031

CNN-based Text-independent Automatic Speaker Identification Using Short Utterances

Mandana Fasounaki, Emirhan Burak Yüce, Serkan Öncül, Gökhan Ince

Department of Computer Engineering

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

3 Citations (Scopus)

Abstract

With the widespread use of voice-controlling services and devices, the research for developing robust and fast systems for automatic speaker identification had accelerated. In this paper, we present a Convolutional Neural Network (CNN) architecture for text-independent automatic speaker identification. The primary purpose is to identify a speaker, among many others, using a short speech segment. Most of the current researches focus on deep CNNs, which were initially designed for computer vision tasks. Besides, most of the existing speaker identification methods require audio samples longer than 3 seconds in the query phase for achieving a high accuracy. We created a CNN architecture appropriate for voice and speech-related classification tasks. We propose an optimum model that achieves 99.5% accuracy on LibriSpeech and 90% accuracy on VoxCeleb 1 dataset using only 1-second test utterances in our experiments.

Original language	English
Title of host publication	Proceedings - 6th International Conference on Computer Science and Engineering, UBMK 2021
Publisher	Institute of Electrical and Electronics Engineers Inc.
Pages	413-418
Number of pages	6
ISBN (Electronic)	9781665429085
DOIs	https://doi.org/10.1109/UBMK52708.2021.9559031
Publication status	Published - 2021
Event	6th International Conference on Computer Science and Engineering, UBMK 2021 - Ankara, Turkey Duration: 15 Sept 2021 → 17 Sept 2021

Publication series

Name	Proceedings - 6th International Conference on Computer Science and Engineering, UBMK 2021

Conference

Conference	6th International Conference on Computer Science and Engineering, UBMK 2021
Country/Territory	Turkey
City	Ankara
Period	15/09/21 → 17/09/21

Bibliographical note

Publisher Copyright:
© 2021 IEEE

Funding

This work has been supported by Arcelik ITU R&D Center and Scientific Project Unit (BAP) of Istanbul Technical University, project number: MOA-2019-42321. The authors thank to Cagri Aslanbas, Berna Erden, Pinar Baki, Ugur Halatoglu and Baris Bayram for their fruitful discussions.

Funders	Funder number
Arcelik ITU R&D Center and Scientific Project Unit
British Association for Psychopharmacology
Istanbul Teknik Üniversitesi	MOA-2019-42321

Keywords

Convolutional Neural Networks
Deep Learning
Signal Processing
Speaker Identifìcation

Access to Document

10.1109/UBMK52708.2021.9559031

Cite this

Fasounaki, M., Yüce, E. B., Öncül, S., & Ince, G. (2021). CNN-based Text-independent Automatic Speaker Identification Using Short Utterances. In Proceedings - 6th International Conference on Computer Science and Engineering, UBMK 2021 (pp. 413-418). (Proceedings - 6th International Conference on Computer Science and Engineering, UBMK 2021). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/UBMK52708.2021.9559031

Fasounaki, Mandana ; Yüce, Emirhan Burak ; Öncül, Serkan et al. / CNN-based Text-independent Automatic Speaker Identification Using Short Utterances. Proceedings - 6th International Conference on Computer Science and Engineering, UBMK 2021. Institute of Electrical and Electronics Engineers Inc., 2021. pp. 413-418 (Proceedings - 6th International Conference on Computer Science and Engineering, UBMK 2021).

@inproceedings{2d66ecf2b5984196b40a025e35dbf1f4,

title = "CNN-based Text-independent Automatic Speaker Identification Using Short Utterances",

abstract = "With the widespread use of voice-controlling services and devices, the research for developing robust and fast systems for automatic speaker identification had accelerated. In this paper, we present a Convolutional Neural Network (CNN) architecture for text-independent automatic speaker identification. The primary purpose is to identify a speaker, among many others, using a short speech segment. Most of the current researches focus on deep CNNs, which were initially designed for computer vision tasks. Besides, most of the existing speaker identification methods require audio samples longer than 3 seconds in the query phase for achieving a high accuracy. We created a CNN architecture appropriate for voice and speech-related classification tasks. We propose an optimum model that achieves 99.5% accuracy on LibriSpeech and 90% accuracy on VoxCeleb 1 dataset using only 1-second test utterances in our experiments.",

keywords = "Convolutional Neural Networks, Deep Learning, Signal Processing, Speaker Identif{\`i}cation",

author = "Mandana Fasounaki and Y{\"u}ce, {Emirhan Burak} and Serkan {\"O}nc{\"u}l and G{\"o}khan Ince",

note = "Publisher Copyright: {\textcopyright} 2021 IEEE; 6th International Conference on Computer Science and Engineering, UBMK 2021 ; Conference date: 15-09-2021 Through 17-09-2021",

year = "2021",

doi = "10.1109/UBMK52708.2021.9559031",

language = "English",

series = "Proceedings - 6th International Conference on Computer Science and Engineering, UBMK 2021",

publisher = "Institute of Electrical and Electronics Engineers Inc.",

pages = "413--418",

booktitle = "Proceedings - 6th International Conference on Computer Science and Engineering, UBMK 2021",

address = "United States",

}

Fasounaki, M, Yüce, EB, Öncül, S & Ince, G 2021, CNN-based Text-independent Automatic Speaker Identification Using Short Utterances. in Proceedings - 6th International Conference on Computer Science and Engineering, UBMK 2021. Proceedings - 6th International Conference on Computer Science and Engineering, UBMK 2021, Institute of Electrical and Electronics Engineers Inc., pp. 413-418, 6th International Conference on Computer Science and Engineering, UBMK 2021, Ankara, Turkey, 15/09/21. https://doi.org/10.1109/UBMK52708.2021.9559031

CNN-based Text-independent Automatic Speaker Identification Using Short Utterances. / Fasounaki, Mandana; Yüce, Emirhan Burak; Öncül, Serkan et al.
Proceedings - 6th International Conference on Computer Science and Engineering, UBMK 2021. Institute of Electrical and Electronics Engineers Inc., 2021. p. 413-418 (Proceedings - 6th International Conference on Computer Science and Engineering, UBMK 2021).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - CNN-based Text-independent Automatic Speaker Identification Using Short Utterances

AU - Fasounaki, Mandana

AU - Yüce, Emirhan Burak

AU - Öncül, Serkan

AU - Ince, Gökhan

PY - 2021

Y1 - 2021

N2 - With the widespread use of voice-controlling services and devices, the research for developing robust and fast systems for automatic speaker identification had accelerated. In this paper, we present a Convolutional Neural Network (CNN) architecture for text-independent automatic speaker identification. The primary purpose is to identify a speaker, among many others, using a short speech segment. Most of the current researches focus on deep CNNs, which were initially designed for computer vision tasks. Besides, most of the existing speaker identification methods require audio samples longer than 3 seconds in the query phase for achieving a high accuracy. We created a CNN architecture appropriate for voice and speech-related classification tasks. We propose an optimum model that achieves 99.5% accuracy on LibriSpeech and 90% accuracy on VoxCeleb 1 dataset using only 1-second test utterances in our experiments.

AB - With the widespread use of voice-controlling services and devices, the research for developing robust and fast systems for automatic speaker identification had accelerated. In this paper, we present a Convolutional Neural Network (CNN) architecture for text-independent automatic speaker identification. The primary purpose is to identify a speaker, among many others, using a short speech segment. Most of the current researches focus on deep CNNs, which were initially designed for computer vision tasks. Besides, most of the existing speaker identification methods require audio samples longer than 3 seconds in the query phase for achieving a high accuracy. We created a CNN architecture appropriate for voice and speech-related classification tasks. We propose an optimum model that achieves 99.5% accuracy on LibriSpeech and 90% accuracy on VoxCeleb 1 dataset using only 1-second test utterances in our experiments.

KW - Convolutional Neural Networks

KW - Deep Learning

KW - Signal Processing

KW - Speaker Identifìcation

UR - http://www.scopus.com/inward/record.url?scp=85125865768&partnerID=8YFLogxK

U2 - 10.1109/UBMK52708.2021.9559031

DO - 10.1109/UBMK52708.2021.9559031

M3 - Conference contribution

AN - SCOPUS:85125865768

T3 - Proceedings - 6th International Conference on Computer Science and Engineering, UBMK 2021

SP - 413

EP - 418

BT - Proceedings - 6th International Conference on Computer Science and Engineering, UBMK 2021

PB - Institute of Electrical and Electronics Engineers Inc.

T2 - 6th International Conference on Computer Science and Engineering, UBMK 2021

Y2 - 15 September 2021 through 17 September 2021

ER -

Fasounaki M, Yüce EB, Öncül S, Ince G. CNN-based Text-independent Automatic Speaker Identification Using Short Utterances. In Proceedings - 6th International Conference on Computer Science and Engineering, UBMK 2021. Institute of Electrical and Electronics Engineers Inc. 2021. p. 413-418. (Proceedings - 6th International Conference on Computer Science and Engineering, UBMK 2021). doi: 10.1109/UBMK52708.2021.9559031

CNN-based Text-independent Automatic Speaker Identification Using Short Utterances

Abstract

Publication series

Conference

Bibliographical note

Funding

Keywords

Access to Document

Other files and links

Fingerprint

Cite this