An approach to classifying proper names for the dispatcher command recognition system
https://doi.org/10.46973/0201-727X_2023_1_175
Abstract
Currently, the spread and actualization of the use of machine learning systems are opening up new horizons for the use of speech recognition systems in all areas of industry, including railway transport, where particularly high requirements for passenger safety are imposed. The authors conducted a review of current patents of existing speech recognition systems, highlighted possible links between patents. The paper also discusses the possible use of transfer learning to improve the speaker's speech recognition performance. Based on the analysis of the literature, the research goal is formulated as the development of a noise-resistant and speaker-independent command recognition algorithm. Within the framework of this paper, the solution of the problem of classifying proper names using low-frequency kepstral coefficients and a convolutional neural network is considered. The authors present an analysis of the results of neural network training on a test sample for a different number of low-frequency kepstral coefficients.
About the Authors
V. G. Sidorenko
Russian University of Transport (MIIT)
Russian Federation
Russian Federation
Sidorenko Valentina Gennadievna, Chair «Control and Information Security», Doctor of Engineering Sciences, Professor
E. P. Balakina
Russian University of Transport (MIIT)
Russian Federation
Russian Federation
Balakina Ekaterina Petrovna, Chair «Control and Information Security», Candidate of Engineering Sciences, Associate Professor
L. N. Loginova
Russian University of Transport (MIIT)
Russian Federation
Russian Federation
Loginova Lyudmila Nikolayevna, Chair «Control and Information Security», Candidate of Engineering Sciences, Associate Professor
M. A. Kulagin
Russian University of Transport (MIIT)
Russian Federation
Russian Federation
Kulagin Maxim Alekseyevich, Chair «Control and Information Security», Candidate of Engineering Sciences, Associate Professor
References
1. Vasiliev, A. S. Analysis of patents as a factor in the study of the technical level of development of technology on the example of jaw crushers / A. S. Vasiliev, N. S. Krupko // Engineering Bulletin of the Don. – 2016. – No. 2. – URL: https://cyberleninka.ru/article/n/analiz-patentov-kak-faktor-issledovaniya-tehnicheskogourovnya-razvitiya-tehniki-na-primere-schekovyh-drobilok (date of access: 05.27.2022).
2. Patent No. 2216052C2 Russian Federation, IPC G10L 15/22, G10L 15/02(2006.01), G10L 15/26. Automatic speech recognition / D. Merrill. – No. 2001104348/09 ; declaration 06.17.1999 ; publ. 11.10.2003.
3. Patent No. WO2013002674A1 International Bureau, IPC G10L 15/187 (2013.01), G10L 25/78 (2013.01). System and method of speech recognition / D. A. Kocharov, A. B. Khomyakov. – Declaration 12.05.2012 ; publ. 03.01.2013 ; Bull. No. 1. – P. 35.
4. Patent No. 2382399С2 Russian Federation, IPC G06F 17/28 (2006.01). Adaptive machine translation / S. D. Richardson, R. F. Rashid. – No. 2004118671/09 ; declaration 06.18.2004 ; publ. 02/20/2010, Bull. No. 5. – P. 36.
5. Patent No. 2628202 C1 Russian Federation, IPC G06F 17/28 (2006.01). Adaptive contextthematic machine translation / M. M. Goldreer. – No. 2016113939 ; declaration 04.11.2016 ; publ. 08.15.2017, Bull. No. 23. – P. 13.
6. Patent No. 2606566С2 Russian Federation, IPC G10L 15/08 (2006.01), G10L 15/00 (2013.01). Method and device for classifying segments of noisy speech using polyspectral analysis / O. N. Titov, A. A. Afanasiev, M. V. Ilyushin. – No. 2014154081 ; declaration 12.29.2014 ; publ. 07.20.2016, Bull. No. 1. – P. 3.
7. Patent No. 2698773С2 Russian Federation, IPC G10L 15/07 (2013.01), G10L 15/22 (2006.01), G10L 15/28 (2013.01). Device and method of speech recognition / K. Arndt Habil, W. Goossen, F. Stefan. – No. 2015118431 ; declaration May 18, 2015 ; publ. 12.10.2016, Bull. No. 34. – Р. 2.
8. Speech technologies in training operational personnel of urban rail transport systems / Balakina, E. P., Kulagin, M. A., Sidorenko, V. G., Loginova, L. N. // Quality. Innovation. Education. – 2022. – No. 3 (179). – P. 36–48. – DOI 10.31145/1999-513x-2022-3-36-48.
9. Chu, Chzhn. Technical description of multilingual and interlingual speech recognition / Chzhn Chu. – URL: https://www.21ic.com/article/828871.html (date of access: 07.29.2022).
10. Stanford Vision and Learning Lab. CS231n Convolutional Neural Networks for Visual Recognition. – URL: https://cs231n.github.io/transferlearning (date of access: 07.30.2022).
11. Security of the use of speech technologies in the work of operational personnel of urban rail transport systems / E. P. Balakina, M. A. Kulagin, L. N. Loginova, V. G. Sidorenko // Problems of managing the safety of complex systems: Proceedings of the XXIX International Scientific and practical conference, Moscow, December 15, 2021. – Moscow : Institute of Management Problems. V. A. Trapeznikova RAN, 2021. – P. 355– 361. – DOI 10.25728/iccss.2021.94.35.056.
12. Ivanov, I. I. Analysis of the method of chalkfrequency cepstral coefficients in relation to the procedure of voice authentication // Actual problems of the humanities and natural sciences. – 2015. – No. 10-1. – URL: https://cyberleninka.ru/article/n/analiz-metoda-melchastotnyh-kepstralnyh-koeffitsientov-primenitelno-k-protsedure-golosovoy-autentifikatsii (date of access: 02.06.2023).
13. Mitchell, T. Machine Learning / T. Mitchell. – Redmond: McGraw-Hill Science/Engineering/Math, 1997. – 432 p. – ISBN 0070428077.
Review
For citations:
Sidorenko V.G., Balakina E.P., Loginova L.N., Kulagin M.A. An approach to classifying proper names for the dispatcher command recognition system. Vestnik Rostovskogo Gosudarstvennogo Universiteta Putej Soobshcheniya. 2023;(1):175-183. (In Russ.) https://doi.org/10.46973/0201-727X_2023_1_175
Views:
4
JATS XML
Email this article 