350 rub
Journal Neurocomputers №2 for 2018 г.
Article in number:
Methodological aspects of automatic speech recognition on the basis of multivariate statistical theory
Type of article: scientific article
UDC: 519.22(075.8)
Keywords: methods speech recognition statistics neural networks acoustic signal spectrum brain posterior probability Markov model computational procedures training.
Authors:

I.V. Shostak − Dr.Sc. (Eng.), Professor, Software Engineering Department, Zhukovsky National Aerospace University «Khakov Aviation Institute» (Kharkov, Ukraine) 

E-mail: iv_shostak@rambler.ru

I.I. Butenko − Ph.D. (Eng.), Assistant Professor, Bauman Moscow State Technical University

E-mail: iuliiabutenko2015@yandex.ru

Abstract:

The article considers the modern methodological approaches of automatic speech recognition with the use of multivariate statistical theory and neural networks. Significant characteristics of the acoustic signal are defined. Distinctive features of the «language» of the computer are studied. It is developed a structural scheme of automatic speech recognition where the speech signal is represented as a random patters. It is chosen optimality criteria for the training data amount and computational procedures requirements to generalize data from the training sample. It is suggested neural network technologies which minimize the objective function between the desired and the generated network output. Spectral characteristics of the speech units are studied.

Pages: 23-33
References
  1. Ivanov V.V. Lingvistika tret'ego tysjacheletija: Voprosy k budushhemu. M.: Jazyki slavjanskoj kul'tury. 2004. 208 s. 
  2. Rassel S., Norvig P. Iskusstvennyj intellekt: sovremennyj podhod. Izd. 2-e / Per. s angl. M.: Izdatel'skij dom «Vil'jams». 2006. 1408 s.
  3. Ljuger D.F. Iskusstvennyj intellekt: strategii i metody reshenija slozhnyh problem. M.: Izdatel'skij dom «Vil'-jams». 2003. 864 c.
  4. Piotrovskij R.G. Lingvisticheskij avtomat (v issledovanii i nepreryvnom obuchenii). Izd. 2-e, dop. i ispr. SPb: Izd-vo RGPU im. A.I. Gercena. 2008. 256 s.
  5. Shherba L.V. Jazykovaja sistema i rechevaja dejatel'nost'. M.: Editorial URSS. 2004. 432 s.
  6. Voloshin V.G. Komp'juternaja lingvistika. Sumy: Universitetskaja kniga. 2004. 382 s.
  7. Bondarenko M.F., Shabanov-Kushnarenko Ju.P. Mozgopodobnye struktury: Spravochnoe posobie. K.: Naukova dumka. 2011. 460 s. 
  8. Selivanova E.A. Lingvisticheskaja jenciklopedija. Poltava: Dovkilja-K. 2010. 844 s.
  9. Bondarenko V.N., Ade F.G. Iskusstvennyj intellekt: Ucheb. posobie dlja vuzov. Sevastopol': Iz-vo SevNTU. 2002. 615 s.
  10. Sidnjaev N.I. Teorija verojatnostej i matematicheskaja statistika: Ucheb. posobie. M.: Izdatel'stvo Jurajt, ID «Jurajt». 2011. 310 s.
  11. Rutkovskaja D., Pilin'skij M., Rutkovskij L. Nejronnye seti, geneticheskie algoritmy i nechetkie sistemy / Per. s pol'sk. I.D. Rudinskogo. M.: Gorjachaja linija Telekom. 2006. 452 s.
  12. Hajkin S. Nejronnye seti: polnyj kurs. M.: Izdatel'skij dom «Vil'jams». 2006. 1104 c.
  13. Makovkin K.A. Gibridnye modeli: skrytye markovskie modeli i nejronnye seti, ih primenenie v sistemah raspo-znavanija rechi // Modeli, metody, algoritmy i arhitektury sistem raspoznavanija rechi. M.: Vychislitel'nyj centr im. A.A. Dorodnicyna. 2006. S. 40−95.
  14. Sidnjaev N.I. Nejronnaja biovozbudimost' i postroenie funkcional'nyh shem iskusstvennogo nejrona // Nejrokom-p'jutery: razrabotka, primenenie. 2012. № 6. S. 24−28.
  15. Kandrashina E.Ju., Litvinceva L.V., Pospelov D.A. Predstavlenie znanij o vremeni i prostranstve v intellektual'nyh sistemah. M.: Nauka. 1989. 328 s.
Date of receipt: 5 декабря 2017 г.