Genetic algorithm based optimization of partly-hidden markov model structure using discriminative criterion

    Research output: Contribution to journalArticle

    2 Citations (Scopus)

    Abstract

    A discriminative modeling is applied to optimize the structure of a Partly-Hidden Markov Model (PHMM). PHMM was proposed in our previous work to deal with the complicated temporal changes of acoustic features. It can represent observation dependent behaviors in both observations and state transitions. In the formulation of the previous PHMM, we used a common structure for all models. However, it is expected that the optimal structure which gives the best performance differs from category to category. In this paper, we designed a new structure optimization method in which the dependence of the states and the observations of PHMM are optimally defined according to each model using the weighted likelihood-ratio maximization (WLRM) criterion. The WLRM criterion gives high discriminability between the correct category and the incorrect categories. Therefore it gives model structures with good discriminative performance. We define the model structure combination which satisfy the WLRM criterion for any possible structure combinations as the optimal structures. A genetic algorithm is also applied to the adequate approximation of a full search. With results of continuous lecture talk speech recognition, the effectiveness of the proposed structure optimization is shown: it reduced the word errors compared to HMM and PHMM with a common structure for all models.

    Original languageEnglish
    Pages (from-to)939-945
    Number of pages7
    JournalIEICE Transactions on Information and Systems
    VolumeE89-D
    Issue number3
    DOIs
    Publication statusPublished - 2006

    Fingerprint

    Hidden Markov models
    Model structures
    Genetic algorithms
    Speech recognition
    Acoustics

    Keywords

    • Acoustic model
    • Genetic algorithm
    • Hidden Markov model
    • Lecture talk speech recognition
    • Partly-hidden Markov model
    • Weighted likelihood-ratio maximization criterion

    ASJC Scopus subject areas

    • Information Systems
    • Computer Graphics and Computer-Aided Design
    • Software

    Cite this

    @article{650b5b06c5d84b3abeb918ae2c85edc4,
    title = "Genetic algorithm based optimization of partly-hidden markov model structure using discriminative criterion",
    abstract = "A discriminative modeling is applied to optimize the structure of a Partly-Hidden Markov Model (PHMM). PHMM was proposed in our previous work to deal with the complicated temporal changes of acoustic features. It can represent observation dependent behaviors in both observations and state transitions. In the formulation of the previous PHMM, we used a common structure for all models. However, it is expected that the optimal structure which gives the best performance differs from category to category. In this paper, we designed a new structure optimization method in which the dependence of the states and the observations of PHMM are optimally defined according to each model using the weighted likelihood-ratio maximization (WLRM) criterion. The WLRM criterion gives high discriminability between the correct category and the incorrect categories. Therefore it gives model structures with good discriminative performance. We define the model structure combination which satisfy the WLRM criterion for any possible structure combinations as the optimal structures. A genetic algorithm is also applied to the adequate approximation of a full search. With results of continuous lecture talk speech recognition, the effectiveness of the proposed structure optimization is shown: it reduced the word errors compared to HMM and PHMM with a common structure for all models.",
    keywords = "Acoustic model, Genetic algorithm, Hidden Markov model, Lecture talk speech recognition, Partly-hidden Markov model, Weighted likelihood-ratio maximization criterion",
    author = "Tetsuji Ogawa and Tetsunori Kobayashi",
    year = "2006",
    doi = "10.1093/ietisy/e89-d.3.939",
    language = "English",
    volume = "E89-D",
    pages = "939--945",
    journal = "IEICE Transactions on Information and Systems",
    issn = "0916-8532",
    publisher = "Maruzen Co., Ltd/Maruzen Kabushikikaisha",
    number = "3",

    }

    TY - JOUR

    T1 - Genetic algorithm based optimization of partly-hidden markov model structure using discriminative criterion

    AU - Ogawa, Tetsuji

    AU - Kobayashi, Tetsunori

    PY - 2006

    Y1 - 2006

    N2 - A discriminative modeling is applied to optimize the structure of a Partly-Hidden Markov Model (PHMM). PHMM was proposed in our previous work to deal with the complicated temporal changes of acoustic features. It can represent observation dependent behaviors in both observations and state transitions. In the formulation of the previous PHMM, we used a common structure for all models. However, it is expected that the optimal structure which gives the best performance differs from category to category. In this paper, we designed a new structure optimization method in which the dependence of the states and the observations of PHMM are optimally defined according to each model using the weighted likelihood-ratio maximization (WLRM) criterion. The WLRM criterion gives high discriminability between the correct category and the incorrect categories. Therefore it gives model structures with good discriminative performance. We define the model structure combination which satisfy the WLRM criterion for any possible structure combinations as the optimal structures. A genetic algorithm is also applied to the adequate approximation of a full search. With results of continuous lecture talk speech recognition, the effectiveness of the proposed structure optimization is shown: it reduced the word errors compared to HMM and PHMM with a common structure for all models.

    AB - A discriminative modeling is applied to optimize the structure of a Partly-Hidden Markov Model (PHMM). PHMM was proposed in our previous work to deal with the complicated temporal changes of acoustic features. It can represent observation dependent behaviors in both observations and state transitions. In the formulation of the previous PHMM, we used a common structure for all models. However, it is expected that the optimal structure which gives the best performance differs from category to category. In this paper, we designed a new structure optimization method in which the dependence of the states and the observations of PHMM are optimally defined according to each model using the weighted likelihood-ratio maximization (WLRM) criterion. The WLRM criterion gives high discriminability between the correct category and the incorrect categories. Therefore it gives model structures with good discriminative performance. We define the model structure combination which satisfy the WLRM criterion for any possible structure combinations as the optimal structures. A genetic algorithm is also applied to the adequate approximation of a full search. With results of continuous lecture talk speech recognition, the effectiveness of the proposed structure optimization is shown: it reduced the word errors compared to HMM and PHMM with a common structure for all models.

    KW - Acoustic model

    KW - Genetic algorithm

    KW - Hidden Markov model

    KW - Lecture talk speech recognition

    KW - Partly-hidden Markov model

    KW - Weighted likelihood-ratio maximization criterion

    UR - http://www.scopus.com/inward/record.url?scp=33645779096&partnerID=8YFLogxK

    UR - http://www.scopus.com/inward/citedby.url?scp=33645779096&partnerID=8YFLogxK

    U2 - 10.1093/ietisy/e89-d.3.939

    DO - 10.1093/ietisy/e89-d.3.939

    M3 - Article

    AN - SCOPUS:33645779096

    VL - E89-D

    SP - 939

    EP - 945

    JO - IEICE Transactions on Information and Systems

    JF - IEICE Transactions on Information and Systems

    SN - 0916-8532

    IS - 3

    ER -