Partly Hidden Markov Model and its application to speech recognition

Tetsunori Kobayashi, Junko Furuyama, Ken Masumitsu

    Research output: Chapter in Book/Report/Conference proceedingChapter

    9 Citations (Scopus)

    Abstract

    A new pattern matching method, Partly Hidden Markov Model, is proposed and applied to speech recognition. Hidden Markov Model, which is widely used for speech recognition, can deal with only piecewise stationary stochastic process. We solved this problem by introducing the modified second order Markov Model, in which the first state is hidden and the second one is observable. In this model, not only the feature parameter observations but also the state transitions are dependent on the previous feature observation. Therefore, even the complicated transient can be modeled precisely. Some simulational experiments showed the high potential of the proposed model. As the results of word recognition test, the error rate was reduced by 39% compared with normal HMM.

    Original languageEnglish
    Title of host publicationICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings
    PublisherIEEE
    Pages121-124
    Number of pages4
    Volume1
    Publication statusPublished - 1999
    EventProceedings of the 1999 IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP-99) - Phoenix, AZ, USA
    Duration: 1999 Mar 151999 Mar 19

    Other

    OtherProceedings of the 1999 IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP-99)
    CityPhoenix, AZ, USA
    Period99/3/1599/3/19

    Fingerprint

    speech recognition
    Hidden Markov models
    Speech recognition
    Pattern matching
    Random processes
    stochastic processes
    Experiments

    ASJC Scopus subject areas

    • Signal Processing
    • Electrical and Electronic Engineering
    • Acoustics and Ultrasonics

    Cite this

    Kobayashi, T., Furuyama, J., & Masumitsu, K. (1999). Partly Hidden Markov Model and its application to speech recognition. In ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings (Vol. 1, pp. 121-124). IEEE.

    Partly Hidden Markov Model and its application to speech recognition. / Kobayashi, Tetsunori; Furuyama, Junko; Masumitsu, Ken.

    ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings. Vol. 1 IEEE, 1999. p. 121-124.

    Research output: Chapter in Book/Report/Conference proceedingChapter

    Kobayashi, T, Furuyama, J & Masumitsu, K 1999, Partly Hidden Markov Model and its application to speech recognition. in ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings. vol. 1, IEEE, pp. 121-124, Proceedings of the 1999 IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP-99), Phoenix, AZ, USA, 99/3/15.
    Kobayashi T, Furuyama J, Masumitsu K. Partly Hidden Markov Model and its application to speech recognition. In ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings. Vol. 1. IEEE. 1999. p. 121-124
    Kobayashi, Tetsunori ; Furuyama, Junko ; Masumitsu, Ken. / Partly Hidden Markov Model and its application to speech recognition. ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings. Vol. 1 IEEE, 1999. pp. 121-124
    @inbook{5abb8147570f4dc7bd39729f48022967,
    title = "Partly Hidden Markov Model and its application to speech recognition",
    abstract = "A new pattern matching method, Partly Hidden Markov Model, is proposed and applied to speech recognition. Hidden Markov Model, which is widely used for speech recognition, can deal with only piecewise stationary stochastic process. We solved this problem by introducing the modified second order Markov Model, in which the first state is hidden and the second one is observable. In this model, not only the feature parameter observations but also the state transitions are dependent on the previous feature observation. Therefore, even the complicated transient can be modeled precisely. Some simulational experiments showed the high potential of the proposed model. As the results of word recognition test, the error rate was reduced by 39{\%} compared with normal HMM.",
    author = "Tetsunori Kobayashi and Junko Furuyama and Ken Masumitsu",
    year = "1999",
    language = "English",
    volume = "1",
    pages = "121--124",
    booktitle = "ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings",
    publisher = "IEEE",

    }

    TY - CHAP

    T1 - Partly Hidden Markov Model and its application to speech recognition

    AU - Kobayashi, Tetsunori

    AU - Furuyama, Junko

    AU - Masumitsu, Ken

    PY - 1999

    Y1 - 1999

    N2 - A new pattern matching method, Partly Hidden Markov Model, is proposed and applied to speech recognition. Hidden Markov Model, which is widely used for speech recognition, can deal with only piecewise stationary stochastic process. We solved this problem by introducing the modified second order Markov Model, in which the first state is hidden and the second one is observable. In this model, not only the feature parameter observations but also the state transitions are dependent on the previous feature observation. Therefore, even the complicated transient can be modeled precisely. Some simulational experiments showed the high potential of the proposed model. As the results of word recognition test, the error rate was reduced by 39% compared with normal HMM.

    AB - A new pattern matching method, Partly Hidden Markov Model, is proposed and applied to speech recognition. Hidden Markov Model, which is widely used for speech recognition, can deal with only piecewise stationary stochastic process. We solved this problem by introducing the modified second order Markov Model, in which the first state is hidden and the second one is observable. In this model, not only the feature parameter observations but also the state transitions are dependent on the previous feature observation. Therefore, even the complicated transient can be modeled precisely. Some simulational experiments showed the high potential of the proposed model. As the results of word recognition test, the error rate was reduced by 39% compared with normal HMM.

    UR - http://www.scopus.com/inward/record.url?scp=0032649321&partnerID=8YFLogxK

    UR - http://www.scopus.com/inward/citedby.url?scp=0032649321&partnerID=8YFLogxK

    M3 - Chapter

    AN - SCOPUS:0032649321

    VL - 1

    SP - 121

    EP - 124

    BT - ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings

    PB - IEEE

    ER -