РАСПОЗНАВАНИЕ РЕЧИ МЕТОДАМИ СКРЫТЫХ МАРКОВСКИХ МОДЕЛЕЙ В АССОЦИАТИВНОЙ ОСЦИЛЛЯТОРНОЙ СРЕДЕ

  • Published on
    05-Apr-2017

  • View
    226

  • Download
    11

Embed Size (px)

Transcript

  • 3 (27), 2013 . ,

    Engineering sciences. Computer science, computer engineering and control 115

    681.3 . . , . .

    . . , -. -, , . . - . - - - , - , - . . - . - Matlab , - - . . - . : , , - .

    I. V. Ognev, P. A. Paramonov

    SPEECH RECOGNITION BY MEANS OF HIDDEN MARKOV MODELS IN ASSOCIATIVE OSCILLOMETRIC MEDIUM

    Abstract. Background. Application of hidden Markov models is based on recursive procedures featuring computational complexity. Herewith, the systems of automatic speech recognition are often required to function in real time mode, and therefore the increase of operation speed thereof is a topical problem. aterials and methods. One of the approaches to solve the said problem is the realization of hardware sup-port of computing in associative oscillometric medium. The said approach is charac-terized by low hardware expenditures due to the simplicity of basic cellular assem-blies and functions performed thereof, as well as by high operation speed independ-ent of the length of the sequence under analysis and of the number of conditions of hidden Markov models, due to concurrency and conveyor nature of computing. Re-sults. The authors suggest hardware implementation to compute the probability function of direct distribution in the medium. The researchers built a program model via Mathlab package in order to experimentally evaluate the precision of computing results in associative oscillometric medium by the example of Russian words recog-nition. Conclusions. The obtained precision value of the results by the example of Russian words recognition demonstrates the efficiency of the applied model. Key words: associative media, speech recognition, hidden markov models.

  • .

    University proceedings. Volga region 116

    () -

    , - , [1, 2]. - ,

    2( )O T N N - T. , - , , . - () , - , - . , , - .

    - . , [3] , , . - , , . - .

    1. : , -

    (. 1). - , - . - , (- , ), (, ), - - [4, 5]. T

    1 2( , , , )TO o o o= . 1 2( , , , )NW w w w= . - : ,W - X [1, 2].

    , - O W W . ( , )h W O W . - W , - :

  • 3 (27), 2013 . ,

    Engineering sciences. Computer science, computer engineering and control 117

    * ( ( , ), )W WW ArgMin d h W O= , (1)

    '( , )d O O 'O O .

    . 1.

    , .W

    - , - .

    40 - .

    1.

    ( , , ) = A B , A - , B , . -. A ija i j. B ( )i kb o - i ko . ,

    i i- . ,

    . , . ( , ) [1, 2, 6, 7].

    1. , .

    2. , (. 2). , , .

    3. , - -. . , , - :

  • .

    University proceedings. Volga region 118

    1) , - . : (, 43), , - , . , , : - ;

    2) - . : , - n-a-z-a-t. // , - : n-a+z z-a+t. , - , .

    4. . .

    5. , - (, ), . - .

    . 2. /n/-/a/,

    : ,

    , - 1 2( , , , )tO o o o= , j t ( , , ) = A B [5, 6]:

    ( )1 1( ) j jj b o = ,

    ( ) ( )11

    ( )sN

    t t ij j ti

    j i a b o=

    = . (2)

    ( )t j . - , .. t T= , ( )T j -, 1 2( , , , )TO o o o= :

  • 3 (27), 2013 . ,

    Engineering sciences. Computer science, computer engineering and control 119

    ( )1

    | ( )sN

    Tj

    P O j=

    = . (3)

    - : k , , O :

    ( )* ArgMax |k

    w P O= . (4)

    ( )|P O - - ( ). ,

    1 2( , , , )TQ q q q= , -, , - t . , - - .

    2.

    () - , -. , - [8].

    -, [810]. , , -. , , , .

    , , , -, . Nq q. PN - [8]:

    Nq

    qPN

    = . (5)

    - PN P :

    lim NN

    P P

    = . (6)

  • .

    University proceedings. Volga region 120

    , : -, (k + 1)- k- ; -, - .

    , . - q, . , , , .

    (. 1).

    1

    ,

    q(k) s(k) o(k + 1)

    o q s q sP P P P P= + 0 0 0 0 1 1 1 0 1 1 1 1

    q(k) s(k) o(k + 1)

    o q s q sP P P P P= + 0 0 0 0 1 1 1 0 1 1 1 1

    q(k) s(k) o(k + 1)

    o q sP P P= 0 0 0 0 1 0 1 0 0 1 1 1

    . . - . - . - .

    -

    , - .

    3.

    , - - . i-

  • 3 (27), 2013 . ,

    Engineering sciences. Computer science, computer engineering and control 121

    i- k- , - (k + 1)- . - { }ija , ( ){ }j kb o { }j , , , .

    , - . - - . , (. 1). - q sP P , .

    . 3 j- t (2).

    . 3. j- t

    , .

    .

    A, B . - . . 4 - m ( )t j . ( )t j , -

  • .

    University proceedings. Volga region 122

    . , , , Ns (- ) - T ( ), - , .

    . 4. m ( )t j

    4.

    - Matlab. - - ( | )P O , , - ( | )sP O :

  • 3 (27), 2013 . ,

    Engineering sciences. Computer science, computer engineering and control 123

    ( | )

    1( | )

    sP OP O

    =

    . (7)

    ( | )P O :

    T; Nq. 20

    , (. 5).

    . 5.

    -

    , . - , M = 28 = 256. , 1 2( , , , )TO o o o= , to - .

    . 6 ( | )P O T. - Nq = 1000 ( ) Nq = 10000 ( -), T 24 36. , T . - , T.

    , ( | )P O , -

    , , , . - 20 80 . . 2.

    , , . -

  • .

    University proceedings. Volga region 124

    - , - , .

    . 6. ( | )P O

    T Nq = 1000 ( ) Nq = 10000 ( )

    2

    , %

    ( | )P O ( | )P O , Nq = 1000

    ( | )P O , Nq = 10000

    95 79 91 Matlab

    -, - .

    1. Becchett i , C. Speech Recognition. Theory and C++ Implementation / C. Becchetti,

    L. P. Ricotti. Wiley, 1999. 428 p. 2. Huang, X . Spoken language processing: a guide to theory, algorithm, and system

    development / X. Huang, A. Acero. Prentice Hall, 2001. 1008 p. 3. Mosleh, M. FPGA implementation of a linear systolic array for speech recognition

    based on HMM / M. Mosleh, S. Setayeshi, M. Mehdi Lotfinejad, A. Mirshekari // The

  • 3 (27), 2013 . ,

    Engineering sciences. Computer science, computer engineering and control 125

    2nd International Conference on Computer and Automation Engineering (ICCAE). 2010. Vol. 3. P. 7578.

    4. , . . / . . , . . // - : . XX . .-. . . : , 2012. . 5358.

    5. Ognev, I . V. The use of extrema distribution as a feature vector for speech patterns recognition / I. V. Ognev, A. I. Ognev, P. A. Paramonov, N. A. Sutula // Pattern Recognition and Image Analysis: New Information Technologies : the 11th Internation-al Conference. 2013. Vol. 1. P. 114117.

    6. Rabiner, L. Fundamentals of speech recognition / L. Rabiner, B.-H. Juang. Pren-tice Hall, 1993. 507 p.

    7. , . - : / . // - . . : , 1989. . 77, 2. . 86120.

    8. , . . : . . . / . . . : (), 2002. 194 .

    9. , . . / . . , . . , . . // : . . . . . 5 (30). : .-. , 2006. 200 .

    10. , . . / . . , . . // . . . . 2006. 6. . 5566.

    References 1. Becchetti C., Ricotti L. P. Speech Recognition. Theory and C++ Implementation.

    Wiley, 1999, 428 p. 2. Huang X., Acero A. Spoken language processing: a guide to theory, algorithm, and

    system de-velopment. Prentice Hall, 2001, 1008 p. 3. Mosleh M., Setayeshi S., Mehdi Lotfinejad M., Mirshekari A. The 2nd International

    Conference on Computer and Automation Engineering (ICCAE). 2010, vol. 3, pp. 7578.

    4. Ognev I. V., Paramonov P. A. Informatsionnye sredstva i tekhnologii: tr. XX Mezhdu-nar. nauch.-tekhn. konf. [Information devices and technology: Proceedings of XXth In-ternational scientific technical conference]. Moscow: MEI, 2012, pp. 5358.

    5. Ognev I. V., Ognev A. I., Paramonov P. A., Sutula N. A. Pattern Recognition and Im-age Analysis: New Information Technologies: the 11th International Conference. 2013, vol. 1, pp. 114117.

    6. Rabiner L., Juang B.-H. Fundamentals of speech recognition. Prentice Hall, 1993, 507 p.

    7. Rabiner L. Trudy instituta inzhenerov po elektrotekhnike i radioelektronike [Proceed-ings of the Institute of electrical engineering and radio electronics]. Moscow: Mir, 1989, vol. 77, no. 2, pp. 86120.

    8. Komarov A. N. Issledovanie i razrabotka assotsiativnykh sred i metodov obrabotki informatsii: dis. kand. tekhn. nauk [Research and development of associative media and methods of data processing: dissertation to apply for the degree of the candidate of en-gineering sciences]. Moscow: MEI(TU), 2002, 194 p.

    9. Komarov A. N., Ognev I. V., Podolin P. B. Vychislitel'nye sistemy i tekhnologii obrabotki informatsii: mezhvuz. sb. nauchn. tr. Vyp. 5 (30) [Computing systems and

  • .

    University proceedings. Volga region 126

    technologies of data processing: interuniversity collected papers. Issue 5 (30)]. Penza: Inf.-izd. tsentr PGU, 2006, 200 p.

    10. Ognev I. V., Podolin P. B. Izvestiya vysshikh uchebnykh zavedeniy. Povolzhskiy region. Tekhnicheskie nauki. [University proceedings. Volga region. Engineering sciences]. 2006, no. 6, pp. 5567.

    , , , (, . , . , 14)

    Ognev Ivan Vasil'evich Doctor of engineering sciences, professor, sub-department of computing technology, National Research University "Moscow Power Engineering University" (14 Krasnokazarmennaya street, Moscow, Russia)

    E-mail: OgnevIV@mpei.ru , (, . , . , 14)

    Paramonov Pavel Aleksandrovich Postgraduate student, National Research University "Moscow Power Engineering University" (14 Krasnokazarmennaya street, Moscow, Russia)

    E-mail: pa.pawka@gmail.com

    681.3

    , . . -

    / . . , . . // - . . . 2013. 3 (27). . 115126.

Recommended

View more >