Feature selection for pose invariant lip biometrics

Adrian Pass, Jianguo Zhang, Darryl Stewart

    Research output: Chapter in Book/Report/Conference proceedingConference contribution

    Abstract

    For the first time in this paper we present results showing the effect of out of plane speaker head pose variation on a lip based speaker verification system. Using appearance DCT based features, we adopt a Mutual Information analysis technique to highlight the class discriminant DCT components most robust to changes in out of plane pose. Experiments are conducted using the initial phase of a new multi view Audio-Visual database designed for research and development of pose-invariant speech and speaker recognition. We show that verification performance can be improved by substituting higher order horizontal DCT components for vertical, particularly in the case of a train/test pose angle mismatch. We further show that the best performance can be achieved by combining this alternative feature selection with multi view training, reporting a relative 45% Equal Error Rate reduction over a common energy based selection. © 2010 ISCA.
    Original languageEnglish
    Title of host publication11th Annual Conference of the International Speech Communication Association 2010 (INTERSPEECH 2010)
    Subtitle of host publicationProceedings
    PublisherInternational Speech Communication Association
    Pages1165-1168
    Number of pages4
    Volume2
    ISBN (Print)9781617821233
    Publication statusPublished - 2010
    Event11th Annual Conference of the International Speech Communication Association: Spoken Language Processing for All - International Conference Hall Makuhari Messe International Convention Complex , Makuhari, Chiba, Japan
    Duration: 26 Sep 201030 Sep 2010
    http://mine.t.u-tokyo.ac.jp/is2010/index.html

    Conference

    Conference11th Annual Conference of the International Speech Communication Association: Spoken Language Processing for All
    Abbreviated titleINTERSPEECH 2010
    CountryJapan
    CityMakuhari, Chiba
    Period26/09/1030/09/10
    Internet address

    Fingerprint

    Biometrics
    Feature extraction
    Information analysis
    Experiments

    Cite this

    Pass, A., Zhang, J., & Stewart, D. (2010). Feature selection for pose invariant lip biometrics. In 11th Annual Conference of the International Speech Communication Association 2010 (INTERSPEECH 2010): Proceedings (Vol. 2, pp. 1165-1168). International Speech Communication Association .
    Pass, Adrian ; Zhang, Jianguo ; Stewart, Darryl. / Feature selection for pose invariant lip biometrics. 11th Annual Conference of the International Speech Communication Association 2010 (INTERSPEECH 2010): Proceedings. Vol. 2 International Speech Communication Association , 2010. pp. 1165-1168
    @inproceedings{082411c0aa8b43989e2fba736c0697af,
    title = "Feature selection for pose invariant lip biometrics",
    abstract = "For the first time in this paper we present results showing the effect of out of plane speaker head pose variation on a lip based speaker verification system. Using appearance DCT based features, we adopt a Mutual Information analysis technique to highlight the class discriminant DCT components most robust to changes in out of plane pose. Experiments are conducted using the initial phase of a new multi view Audio-Visual database designed for research and development of pose-invariant speech and speaker recognition. We show that verification performance can be improved by substituting higher order horizontal DCT components for vertical, particularly in the case of a train/test pose angle mismatch. We further show that the best performance can be achieved by combining this alternative feature selection with multi view training, reporting a relative 45{\%} Equal Error Rate reduction over a common energy based selection. {\circledC} 2010 ISCA.",
    author = "Adrian Pass and Jianguo Zhang and Darryl Stewart",
    year = "2010",
    language = "English",
    isbn = "9781617821233",
    volume = "2",
    pages = "1165--1168",
    booktitle = "11th Annual Conference of the International Speech Communication Association 2010 (INTERSPEECH 2010)",
    publisher = "International Speech Communication Association",

    }

    Pass, A, Zhang, J & Stewart, D 2010, Feature selection for pose invariant lip biometrics. in 11th Annual Conference of the International Speech Communication Association 2010 (INTERSPEECH 2010): Proceedings. vol. 2, International Speech Communication Association , pp. 1165-1168, 11th Annual Conference of the International Speech Communication Association: Spoken Language Processing for All, Makuhari, Chiba, Japan, 26/09/10.

    Feature selection for pose invariant lip biometrics. / Pass, Adrian; Zhang, Jianguo; Stewart, Darryl.

    11th Annual Conference of the International Speech Communication Association 2010 (INTERSPEECH 2010): Proceedings. Vol. 2 International Speech Communication Association , 2010. p. 1165-1168.

    Research output: Chapter in Book/Report/Conference proceedingConference contribution

    TY - GEN

    T1 - Feature selection for pose invariant lip biometrics

    AU - Pass, Adrian

    AU - Zhang, Jianguo

    AU - Stewart, Darryl

    PY - 2010

    Y1 - 2010

    N2 - For the first time in this paper we present results showing the effect of out of plane speaker head pose variation on a lip based speaker verification system. Using appearance DCT based features, we adopt a Mutual Information analysis technique to highlight the class discriminant DCT components most robust to changes in out of plane pose. Experiments are conducted using the initial phase of a new multi view Audio-Visual database designed for research and development of pose-invariant speech and speaker recognition. We show that verification performance can be improved by substituting higher order horizontal DCT components for vertical, particularly in the case of a train/test pose angle mismatch. We further show that the best performance can be achieved by combining this alternative feature selection with multi view training, reporting a relative 45% Equal Error Rate reduction over a common energy based selection. © 2010 ISCA.

    AB - For the first time in this paper we present results showing the effect of out of plane speaker head pose variation on a lip based speaker verification system. Using appearance DCT based features, we adopt a Mutual Information analysis technique to highlight the class discriminant DCT components most robust to changes in out of plane pose. Experiments are conducted using the initial phase of a new multi view Audio-Visual database designed for research and development of pose-invariant speech and speaker recognition. We show that verification performance can be improved by substituting higher order horizontal DCT components for vertical, particularly in the case of a train/test pose angle mismatch. We further show that the best performance can be achieved by combining this alternative feature selection with multi view training, reporting a relative 45% Equal Error Rate reduction over a common energy based selection. © 2010 ISCA.

    M3 - Conference contribution

    SN - 9781617821233

    VL - 2

    SP - 1165

    EP - 1168

    BT - 11th Annual Conference of the International Speech Communication Association 2010 (INTERSPEECH 2010)

    PB - International Speech Communication Association

    ER -

    Pass A, Zhang J, Stewart D. Feature selection for pose invariant lip biometrics. In 11th Annual Conference of the International Speech Communication Association 2010 (INTERSPEECH 2010): Proceedings. Vol. 2. International Speech Communication Association . 2010. p. 1165-1168