Neurorecognition visualization in multitask end-to-end speech

被引：0

作者：

Mamyrbayev, Orken ^{[1
]}

Pavlov, Sergii ^{[2
]}

Bekarystankyzy, Akbayan ^{[3
,4
]}

Oralbekova, Dina ^{[5
]}

Zhumazhanov, Bagashar ^{[1
]}

Azarova, Larysa

Mussayeva, Dinara ^{[6
]}

Koval, Tetiana ^{[7
]}

Gromaszek, Konrad ^{[8
]}

Issimov, Nurdaulet ^{[9
]}

Shiyapov, Kadrzhan ^{[10
]}

机构：

[1] Inst Informat & Computat Technol, 28 Shevchenko St, Alma Ata 050010, Kazakhstan

[2] Vinnytsia Natl Tech Univ, Khmelnytske Hwy,95, UA-21000 Vinnytsia, Ukraine

[3] Satbayev Univ, Satpaev St 22, Alma Ata 050000, Kazakhstan

[4] Narxoz Univ, Zhandossov St 55, Alma Ata 050035, Kazakhstan

[5] Almaty Univ Power Engn & Telecommun, Baytursynuli St 126-1, Alma Ata 050013, Kazakhstan

[6] Inst Econ CS MES RK, Kurmangazy St 29,A25K1B0, Alma Ata, Kazakhstan

[7] Vinnytsia Mykhailo Kotsiubynskyi State Pedag Univ, Ostrozkoho St 32, UA-21100 Vinnytsia, Ukraine

[8] Lublin Univ Technol, Ul Nadbystrzycka 38D, PL-20618 Lublin, Poland

[9] Turan Univ, Satpayeva St 16a, Alma Ata 050013, Kazakhstan

[10] Abai Kazakh Natl Pedag Univ, Dostyk Ave 13, Alma Ata 050010, Kazakhstan

来源：

OPTICAL FIBERS AND THEIR APPLICATIONS 2023 | 2024年 / 12985卷

关键词：

end-to-end; multitask training; speech recognition; speaker identification; dialect identification; NEURAL-NETWORKS;

D O I：

10.1117/12.3022727

中图分类号：

TM [电工技术]; TN [电子技术、通信技术];

学科分类号：

0808 ; 0809 ;

摘要：

Nowadays, speech-processing technologies with different language systems are successfully used in mobile and stationary devices. Kazakh is considered a low-resource language, which poses various challenges for conventional speech recognition methods. This paper presents a proposed model capable of multitasking and handling concurrent speech recognition, dialect identification, and speaker identification, all in an end-to-end framework. The developed multitask model enables training three different tasks within a single model. A multitask recognition system is created based on the WaveNet-CTC model. Experiments show that for the concrete task end-to-end multitask model has better performance than other models..

引用

页数：8

共 30 条

[1]

Bisikalo O., 2023, Entropy., V25, P1

[2]

Cai L., 2008, Gansu Sci. Technol, V24, P46

[3] Multitask Learning of Deep Neural Networks for Low-Resource Speech Recognition [J].

Chen, Dongpeng ;

Mak, Brian Kan-Wing .

IEEE-ACM TRANSACTIONS ON AUDIO SPEECH AND LANGUAGE PROCESSING, 2015, 23 (07) :1172-1183

[4] Automatic Speech Recognition for Uyghur, Kazakh, and Kyrgyz: An Overview [J].

Du, Wenqiang ;

Maimaitiyiming, Yikeremu ;

Nijat, Mewlude ;

Li, Lantian ;

Hamdulla, Askar ;

Wang, Dong .

APPLIED SCIENCES-BASEL, 2023, 13 (01)

[5]

Han Q., 2010, Softw. Guide., V9, P173

[6] End-to-end Japanese Multi-dialect Speech Recognition and Dialect Identification with Multi-task Learning [J].

Imaizumi, Ryo ;

Masumura, Ryo ;

Shiota, Sayaka ;

Kiya, Hitoshi .

APSIPA TRANSACTIONS ON SIGNAL AND INFORMATION PROCESSING, 2022, 11 (01)

[7]

Khassanov Y, 2021, 16TH CONFERENCE OF THE EUROPEAN CHAPTER OF THE ASSOCIATION FOR COMPUTATIONAL LINGUISTICS (EACL 2021), P697

[8]

Kim S, 2017, INT CONF ACOUST SPEE, P4835, DOI 10.1109/ICASSP.2017.7953075

[9] IDENTIFICATION OF CLADDING MODES IN SMF-28 FIBERS WITH TFBG STRUCTURES [J].

Kisala, Piotr ;

Kalizhanova, Aliy ;

Kozbakova, Ainur ;

Yeraliyeva, Bakhyt .

METROLOGY AND MEASUREMENT SYSTEMS, 2023, 30 (03) :507-518

[10]

Krishna K., 2018, Hierarchical multitask learning for CTC-based speech recognition, P543

← 1 2 3 →