{"id":1996,"date":"2021-01-18T18:26:31","date_gmt":"2021-01-18T16:26:31","guid":{"rendered":"http:\/\/journals.khnu.km.ua\/vestnik\/?p=1996"},"modified":"2021-05-19T13:36:34","modified_gmt":"2021-05-19T10:36:34","slug":"%d0%bd%d0%b5%d0%b9%d1%80%d0%be%d0%bc%d0%b5%d1%80%d0%b5%d0%b6%d0%b5%d0%b2%d0%b0-%d0%b0%d0%b4%d0%b0%d0%bf%d1%82%d0%b0%d1%86%d1%96%d1%8f-plda-%d0%b4%d0%bb%d1%8f-%d0%b2%d0%b8%d0%ba%d0%be%d1%80%d0%b8%d1%81","status":"publish","type":"post","link":"https:\/\/journals.khnu.km.ua\/vestnik\/?p=1996","title":{"rendered":"\u041d\u0435\u0439\u0440\u043e\u043c\u0435\u0440\u0435\u0436\u0435\u0432\u0430 \u0430\u0434\u0430\u043f\u0442\u0430\u0446\u0456\u044f plda \u0434\u043b\u044f \u0432\u0438\u043a\u043e\u0440\u0438\u0441\u0442\u0430\u043d\u043d\u044f \u0443 \u0430\u0432\u0442\u043e\u043c\u0430\u0442\u0438\u0437\u043e\u0432\u0430\u043d\u0456\u0439 \u0441\u0438\u0441\u0442\u0435\u043c\u0456 \u0440\u043e\u0437\u043f\u0456\u0437\u043d\u0430\u0432\u0430\u043d\u043d\u044f \u043c\u043e\u0432\u0446\u044f \u043a\u0440\u0438\u0442\u0438\u0447\u043d\u043e\u0433\u043e \u0437\u0430\u0441\u0442\u043e\u0441\u0443\u0432\u0430\u043d\u043d\u044f"},"content":{"rendered":"

<\/p>\n

\u041d\u0415\u0419\u0420\u041e\u041c\u0415\u0420\u0415\u0416\u0415\u0412\u0410 \u0410\u0414\u0410\u041f\u0422\u0410\u0426\u0406\u042f PLDA \u0414\u041b\u042f \u0412\u0418\u041a\u041e\u0420\u0418\u0421\u0422\u0410\u041d\u041d\u042f \u0423 \u0410\u0412\u0422\u041e\u041c\u0410\u0422\u0418\u0417\u041e\u0412\u0410\u041d\u0406\u0419 \u0421\u0418\u0421\u0422\u0415\u041c\u0406 \u0420\u041e\u0417\u041f\u0406\u0417\u041d\u0410\u0412\u0410\u041d\u041d\u042f \u041c\u041e\u0412\u0426\u042f \u041a\u0420\u0418\u0422\u0418\u0427\u041d\u041e\u0413\u041e \u0417\u0410\u0421\u0422\u041e\u0421\u0423\u0412\u0410\u041d\u041d\u042f<\/p>\n

NEURAL NETWORK ADAPTATION OF PLDA FOR THE AUTOMATIC SPEAKER RECOGNITION SYSTEM OF CRITICAL USE<\/p>\n

<\/a> \u0421\u0442\u043e\u0440\u0456\u043d\u043a\u0438: 172-177. \u041d\u043e\u043c\u0435\u0440: \u21161, 2019 (269)<\/strong>
\n\u0410\u0432\u0442\u043e\u0440\u0438:<\/strong>
\n\u0410.\u0414. \u0413\u0410\u0424\u0423\u0420\u041e\u0412\u0410, \u0412.\u0412. \u041a\u041e\u0412\u0422\u0423\u041d
\n\u0412\u0456\u043d\u043d\u0438\u0446\u044c\u043a\u0438\u0439 \u043d\u0430\u0446\u0456\u043e\u043d\u0430\u043b\u044c\u043d\u0438\u0439 \u0442\u0435\u0445\u043d\u0456\u0447\u043d\u0438\u0439 \u0443\u043d\u0456\u0432\u0435\u0440\u0441\u0438\u0442\u0435\u0442
\nA.D. HAFUROVA, V.V. KOVTUN
\nVinnytsia National Technical University
\nDOI:<\/strong> https:\/\/www.doi.org\/10.31891\/2307-5732-2019-269-1-172-177<\/a>
\n\u0420\u0435\u0446\u0435\u043d\u0437\u0456\u044f\/Peer review :<\/strong> 16.10.2019 \u0440.
\n\u041d\u0430\u0434\u0440\u0443\u043a\u043e\u0432\u0430\u043d\u0430\/Printed :<\/strong> 16.02.2019 \u0440.<\/p>\n

\u0410\u043d\u043e\u0442\u0430\u0446\u0456\u044f \u043c\u043e\u0432\u043e\u044e \u043e\u0440\u0438\u0433\u0456\u043d\u0430\u043b\u0443<\/strong><\/p>\n

\u0410\u0432\u0442\u043e\u0440\u0438 \u043f\u0440\u043e\u043f\u043e\u043d\u0443\u044e\u0442\u044c \u0430\u043a\u0442\u0443\u0430\u043b\u044c\u043d\u0456 \u0441\u0438\u0441\u0442\u0435\u043c\u0438 \u0440\u043e\u0437\u043f\u0456\u0437\u043d\u0430\u0432\u0430\u043d\u043d\u044f \u043c\u043e\u0432\u0446\u0456\u0432, \u0434\u0435 \u0437\u0430\u0441\u0442\u043e\u0441\u043e\u0432\u0443\u0454\u0442\u044c\u0441\u044f i-\u0432\u0435\u043a\u0442\u043e\u0440\u043d\u0435\/PLDA \u043c\u043e\u0434\u0435\u043b\u044e\u0432\u0430\u043d\u043d\u044f \u0434\u043b\u044f \u043e\u043f\u0438\u0441\u0443 \u0444\u043e\u043d\u043e\u0433\u0440\u0430\u043c, \u0441\u0438\u043d\u0442\u0435\u0437\u0443\u044e\u0442\u044c \u0443\u0437\u0430\u0433\u0430\u043b\u044c\u043d\u0435\u043d\u0443 PLDA \u043c\u043e\u0434\u0435\u043b\u044c \u0456\u0437 \u0443\u0441\u0435\u0440\u0435\u0434\u043d\u0435\u043d\u0438\u043c\u0438 \u043f\u0430\u0440\u0430\u043c\u0435\u0442\u0440\u0430\u043c\u0438 \u043f\u043e \u0432\u0441\u0456\u0439 \u0431\u0430\u0437\u0456 \u0444\u043e\u043d\u043e\u0433\u0440\u0430\u043c \u0431\u0435\u0437 \u0457\u0445 \u0441\u0435\u0433\u0440\u0435\u0433\u0430\u0446\u0456\u0457 \u0437\u0430 \u0440\u0456\u0432\u043d\u0435\u043c \u0448\u0443\u043c\u0456\u0432. \u0412 \u0440\u0435\u0437\u0443\u043b\u044c\u0442\u0430\u0442\u0456 \u0442\u0430\u043a\u0456 \u0441\u0438\u0441\u0442\u0435\u043c\u0438 \u0437\u0430\u0431\u0435\u0437\u043f\u0435\u0447\u0443\u044e\u0442\u044c \u043f\u0440\u0438\u0439\u043d\u044f\u0442\u043d\u0438\u0439 \u0440\u0456\u0432\u0435\u043d\u044c \u043d\u0430\u0434\u0456\u0439\u043d\u043e\u0441\u0442\u0456 \u043b\u0438\u0448\u0435 \u0437\u0430 \u043d\u0430\u044f\u0432\u043d\u043e\u0441\u0442\u0456 \u0432\u0435\u043b\u0438\u043a\u043e\u0457 \u043d\u0430\u0432\u0447\u0430\u043b\u044c\u043d\u043e\u0457 \u0432\u0438\u0431\u0456\u0440\u043a\u0438 \u044f\u043a \u0437\u0430 \u043a\u0456\u043b\u044c\u043a\u0456\u0441\u0442\u044e, \u0442\u0430\u043a \u0456 \u0437\u0430 \u0442\u0440\u0438\u0432\u0430\u043b\u0456\u0441\u0442\u044e \u0444\u043e\u043d\u043e\u0433\u0440\u0430\u043c. \u0423 \u0440\u043e\u0431\u043e\u0442\u0456 \u0430\u0432\u0442\u043e\u0440\u0438 \u0441\u0438\u043d\u0442\u0435\u0437\u0443\u0432\u0430\u043b\u0438 \u043e\u043a\u0440\u0435\u043c\u0456 PLDA \u043c\u043e\u0434\u0435\u043b\u0456 \u0434\u043b\u044f \u043e\u043f\u0438\u0441\u0443 \u0444\u043e\u043d\u043e\u0433\u0440\u0430\u043c \u0456\u0437 \u0434\u0435\u0442\u0435\u0440\u043c\u0456\u043d\u043e\u0432\u0430\u043d\u0438\u043c\u0438 \u0440\u0456\u0432\u043d\u044f\u043c\u0438 \u0432\u0456\u0434\u043d\u043e\u0448\u0435\u043d\u043d\u044f \u0441\u0438\u0433\u043d\u0430\u043b\/\u0448\u0443\u043c (\u0412\u0421\u0428), \u0432 \u0440\u0435\u0437\u0443\u043b\u044c\u0442\u0430\u0442\u0456 \u0447\u043e\u0433\u043e \u0444\u0430\u043a\u0442\u043e\u0440\u0438, \u044f\u043a\u0456 \u0445\u0430\u0440\u0430\u043a\u0442\u0435\u0440\u0438\u0437\u0443\u044e\u0442\u044c \u0456\u043d\u0434\u0438\u0432\u0456\u0434\u0443\u0430\u043b\u044c\u043d\u0456\u0441\u0442\u044c \u043c\u043e\u0432\u0446\u0456\u0432, \u0437\u043e\u0441\u0435\u0440\u0435\u0434\u0436\u0435\u043d\u043e \u0443 \u043d\u0430\u0439\u0431\u0456\u043b\u044c\u0448 \u043c\u0456\u043d\u043b\u0438\u0432\u0438\u0445 \u043e\u0431\u043b\u0430\u0441\u0442\u044f\u0445 \u0456-\u0432\u0435\u043a\u0442\u043e\u0440\u043d\u043e\u0433\u043e \u043f\u0440\u043e\u0441\u0442\u043e\u0440\u0443. \u0410\u0432\u0442\u043e\u0440\u0438 \u0437\u0430\u043f\u0440\u043e\u043f\u043e\u043d\u043e\u0432\u0443\u0432\u0430\u043b\u0438 \u0432\u0438\u043a\u043e\u0440\u0438\u0441\u0442\u043e\u0432\u0443\u0432\u0430\u0442\u0438 \u043d\u0435\u0439\u0440\u043e\u043c\u0435\u0440\u0435\u0436\u0443 \u0434\u043b\u044f \u043f\u0440\u0435\u0446\u0438\u0437\u0456\u0439\u043d\u043e\u0433\u043e \u0434\u0435\u0442\u0435\u043a\u0442\u0443\u0432\u0430\u043d\u043d\u044f \u0434\u0456\u0430\u043f\u0430\u0437\u043e\u043d\u0456\u0432 \u0440\u0456\u0432\u043d\u0456\u0432 \u0412\u0421\u0428 \u0432\u0445\u0456\u0434\u043d\u0438\u0445 \u0444\u043e\u043d\u043e\u0433\u0440\u0430\u043c \u0456\u0437 \u043f\u043e\u0434\u0430\u043b\u044c\u0448\u0438\u043c \u0432\u0438\u043a\u043e\u0440\u0438\u0441\u0442\u0430\u043d\u043d\u044f\u043c \u043e\u0434\u0435\u0440\u0436\u0430\u043d\u0438\u0445 \u0434\u0430\u043d\u0438\u0445 \u043f\u0440\u0438 \u0441\u0438\u043d\u0442\u0435\u0437\u0456 \u0443\u043d\u0456\u0432\u0435\u0440\u0441\u0430\u043b\u044c\u043d\u0438\u0445 \u0444\u043e\u043d\u043e\u0432\u0438\u0445 \u043c\u043e\u0434\u0435\u043b\u0435\u0439, \u044f\u043a\u0456 \u043e\u043f\u0442\u0438\u043c\u0430\u043b\u044c\u043d\u043e \u043e\u043f\u0438\u0441\u0443\u044e\u0442\u044c \u0432\u043f\u043b\u0438\u0432 \u0441\u0442\u043e\u0440\u043e\u043d\u043d\u0456\u0445 \u0430\u043a\u0443\u0441\u0442\u0438\u0447\u043d\u0438\u0445 \u0448\u0443\u043c\u0456\u0432 \u0443 \u0444\u043e\u043d\u043e\u0433\u0440\u0430\u043c\u0430\u0445, \u0449\u043e \u0434\u043e\u0437\u0432\u043e\u043b\u044f\u0454 \u044f\u043a \u043f\u0456\u0434\u0432\u0438\u0449\u0438\u0442\u0438 \u044f\u043a\u0456\u0441\u043d\u0456 \u043f\u043e\u043a\u0430\u0437\u043d\u0438\u043a\u0438 \u0440\u043e\u0431\u043e\u0442\u0438 \u0430\u0432\u0442\u043e\u043c\u0430\u0442\u0438\u0437\u043e\u0432\u0430\u043d\u043e\u0457 \u0441\u0438\u0441\u0442\u0435\u043c\u0438 \u0440\u043e\u0437\u043f\u0456\u0437\u043d\u0430\u0432\u0430\u043d\u043d\u044f \u043c\u043e\u0432\u0446\u0456\u0432 \u043a\u0440\u0438\u0442\u0438\u0447\u043d\u043e\u0433\u043e \u0437\u0430\u0441\u0442\u043e\u0441\u0443\u0432\u0430\u043d\u043d\u044f \u0442\u0430\u043a \u0456 \u0432\u0441\u0442\u0430\u043d\u043e\u0432\u043b\u044e\u0432\u0430\u0442\u0438 \u0444\u0430\u043a\u0442 \u043d\u0435\u043f\u0440\u0438\u0434\u0430\u0442\u043d\u043e\u0441\u0442\u0456 \u0434\u043b\u044f \u043f\u043e\u0434\u0430\u043b\u044c\u0448\u043e\u0457 \u043e\u0431\u0440\u043e\u0431\u043a\u0438 \u0437\u0430\u043f\u0440\u043e\u043f\u043e\u043d\u043e\u0432\u0430\u043d\u043e\u0457 \u0441\u0438\u0441\u0442\u0435\u043c\u0456 \u0432 \u044f\u043a\u043e\u0441\u0442\u0456 \u0432\u0445\u0456\u0434\u043d\u0438\u0445 \u0434\u0430\u043d\u0438\u0445 \u0444\u043e\u043d\u043e\u0433\u0440\u0430\u043c\u0438, \u0449\u043e \u0442\u0430\u043a\u043e\u0436 \u043f\u0456\u0434\u0432\u0438\u0449\u0443\u0454 \u043d\u0430\u0434\u0456\u0439\u043d\u0456\u0441\u0442\u044c \u0440\u043e\u0431\u043e\u0442\u0438 \u0441\u0438\u0441\u0442\u0435\u043c\u0438 \u0437\u0430\u0433\u0430\u043b\u043e\u043c.
\n\u041a\u043b\u044e\u0447\u043e\u0432\u0456 \u0441\u043b\u043e\u0432\u0430:<\/strong> \u0430\u0432\u0442\u043e\u043c\u0430\u0442\u0438\u0437\u043e\u0432\u0430\u043d\u0430 \u0441\u0438\u0441\u0442\u0435\u043c\u0430 \u0440\u043e\u0437\u043f\u0456\u0437\u043d\u0430\u0432\u0430\u043d\u043d\u044f \u043c\u043e\u0432\u0446\u0456\u0432 \u043a\u0440\u0438\u0442\u0438\u0447\u043d\u043e\u0433\u043e \u0437\u0430\u0441\u0442\u043e\u0441\u0443\u0432\u0430\u043d\u043d\u044f, \u0456-\u0432\u0435\u043a\u0442\u043e\u0440\u0438, \u043d\u0435\u0439\u0440\u043e\u043c\u0435\u0440\u0435\u0436\u0430, \u0441\u0443\u043c\u0456\u0448 PLDA<\/p>\n

\u0420\u043e\u0437\u0448\u0438\u0440\u0435\u043d\u0430 \u0430\u043d\u043e\u0442\u0430\u0446\u0456\u044f \u0430\u043d\u0433\u043b\u0456\u0439\u0441\u044c\u043a\u043e\u044e \u043c\u043e\u0432\u043e\u044e<\/strong><\/p>\n

.Authors offer speaker recognition systems that use i-vector\/PLDA modelling to describe phonograms synthesize a generalized PLDA model with averaged parameters throughout the phonogram database without their noise segregation. As a result, such systems provide an acceptable level of reliability only in the presence of a large training sample, both in quantity and duration of phonograms. The authors synthesized individual PLDA models for describing phonograms with deterministic levels of signal-to-noise ratio (SNR), resulting in factors that characterize the individuality of speakers, concentrated in the most volatile regions of the i-vector space. Authors proposed to use a neural network for precise detection of ranges of SNR levels of incoming phonograms with subsequent use of the data obtained in the synthesis of universal background models that optimally describe the influence of extraneous acoustic noises in phonograms, which allows not only to improve the performance of the automatic speaker recognition system of critical use, but also to establish the fact of unfitness for further processing of the input phonogram, which also increases the reliability of the system at all. The authors proposed an improved method for the adaptation of mixtures of PLDA-models to the presence of speech signals in phonograms, on which ACCRM performs voice recognition, dynamic level of VSS. The method based on the use of GNM for training UFM, namely, the GNM in the learning process changes the weight of interneuronal bonds to optimally determine the ranges of changes in the level of BCS in the i-vectors extracted from educational phonograms.
\nKeywords:<\/strong> automatic speaker recognition system of critical use, i-vectors, neural network, PLDA mixture.<\/p>\n

References<\/strong><\/p>\n

    \n
  1. Kovtun V.V. Pidvyshchennia shumostiikosti avtomatyzovanoi systemy rozpiznavannia movtsia krytychnoho zastosuvannia \/ T.V. Hryshchuk, V.V. Kovtun \/\/ Visnyk Vinnytskoho politekhnichnoho instytutu. \u2013 2018. \u2013 \u2116 1. \u2013 S. 98\u2013111.<\/li>\n
  2. Reynolds D. A. Speaker verification using adapted Gaussian mixture models \/ D. A. Reynolds, T. F. Quatieri, R. B. Dunn \/\/ Digital Signal Processing. \u2013 2000. \u2013 Vol. 10. \u2013 \u2116 1. \u2013 R. 19\u201341.<\/li>\n
  3. Shao Y. Robust speaker identification using auditory features and computational auditory scene analysis \/ Y. Shao, D. Wang \/\/ IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). \u2013 2008. \u2013 R. 1589\u20131592.<\/li>\n
  4. Elder J.H. Probabilistic linear discriminant analysis for inferences about identity \/ S.J.D. Prince, J.H. Elder \/\/ ICCV. \u2013 2007. \u2013 R. 1\u20138.<\/li>\n
  5. Yaman S. Bottleneck features for speaker recognition \/ S. Yaman, J.W. Pelecanos, R. Sarikaya \/\/ Odyssey. \u2013 2012. \u2013 Vol. 12. \u2013 R. 105\u2013108.<\/li>\n
  6. Ghahabi O. Deep belief networks for i-vector based speaker recognition \/ O. Ghahabi, J. Hernando \/\/ ICASSP. \u2013 2014. \u2013 R. 1700\u20131704.<\/li>\n
  7. Variani E. Deep neural networks for small footprint text-dependent speaker verification \/ E. Variani, X. Lei, E. McDermott, I. Lopez M., J. Gonzalez-Dominguez \/\/ ICASSP. \u2013 2014. \u2013 R. 4052\u20134056.<\/li>\n
  8. Zhao X.J. Deep neural networks for cochannel speaker identification \/ X.J. Zhao, Y.X. Wang, D.L. Wang \/ ICASSP. \u2013 2015. \u2013 R. 4824\u20134828.<\/li>\n
  9. Garcia-Romero D. Multicondition training of Gaussian PLDA models in i-vector space for noise and reverberation robust speaker recognition \/ D. Garcia-Romero, X. Zhou, C.Y. Espy-Wilson \/\/ ICASSP. \u2013 2012. \u2013 R. 4257\u20134260.<\/li>\n
  10. McLachlan G. Mixtures of factor analyzers \/ G. McLachlan, D. Peel \/\/ Finite Mixture Models. \u2013 2000. \u2013 R. 238\u2013256.<\/li>\n
  11. Ghahramani Z. The EM algorithm for mixtures of factor analyzers \/ Z. Ghahramani, G.E. Hinton \/\/ Technical Report CRGTR-96-1, University of Toronto. \u2013 1996.<\/li>\n
  12. Garcia-Romero D. Analysis of i-vector length normalization in speaker recognition systems \/ D. Garcia-Romero, C.Y. Espy-Wilson \/\/ Interspeech. \u2013 2011. \u2013 R. 249\u2013252.<\/li>\n
  13. Kovtun V.V. Optymizatsiia alfavitu informatyvnykh oznak dlia avtomatyzovanoi systemy rozpiznavannia movtsiv krytychnoho zastosuvannia \/ A.O. Bereza, M.M. Bykov, A.D. Hafurova, V.V. Kovtun \/\/ Herald of Khmelnytskyi National University. \u2013 2017. \u2013 \u2116 3(249). \u2013 S. 222\u2013228.<\/li>\n
  14. Kovtun V.V. Vykorystannia mnozhyny mikrofoniv u avtomatyzovanii systemi rozpiznavannia movtsia krytychnoho zastosuvannia \/ M.M. Bykov, V.V. Kovtun \/\/ Visnyk Vinnytskoho politekhnichnoho instytutu, Vinnytsia. \u2013 2017. \u2013 \u2116 3. \u2013 S. 84\u201391.<\/li>\n
  15. Hatch A. Within-class covariance normalization for SVM-based speaker recognition \/ A. Hatch, S. Kajarekar, A. Stolcke \/\/ ISCSLP, Pittsburgh. \u2013 2006. \u2013 R. 1471\u20131474.<\/li>\n
  16. Bengio Y. Learning deep architectures for AI \/ Y. Bengio \/\/ Foundations and trends R in Machine Learning. \u2013 2009. \u2013 Vol. 2. \u2013 \u2116 1. \u2013 R. 1\u2013127.<\/li>\n<\/ol>\n

     <\/p>\n","protected":false},"excerpt":{"rendered":"","protected":false},"author":1,"featured_media":0,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":[],"categories":[21],"tags":[],"_links":{"self":[{"href":"https:\/\/journals.khnu.km.ua\/vestnik\/index.php?rest_route=\/wp\/v2\/posts\/1996"}],"collection":[{"href":"https:\/\/journals.khnu.km.ua\/vestnik\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/journals.khnu.km.ua\/vestnik\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/journals.khnu.km.ua\/vestnik\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/journals.khnu.km.ua\/vestnik\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=1996"}],"version-history":[{"count":5,"href":"https:\/\/journals.khnu.km.ua\/vestnik\/index.php?rest_route=\/wp\/v2\/posts\/1996\/revisions"}],"predecessor-version":[{"id":6438,"href":"https:\/\/journals.khnu.km.ua\/vestnik\/index.php?rest_route=\/wp\/v2\/posts\/1996\/revisions\/6438"}],"wp:attachment":[{"href":"https:\/\/journals.khnu.km.ua\/vestnik\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=1996"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/journals.khnu.km.ua\/vestnik\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=1996"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/journals.khnu.km.ua\/vestnik\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=1996"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}