An Improved NN Training Scheme Using Two-Stage LDA Features for Face Recognition

Seyed Bozorgtabar, Roland Goecke

    Research output: A Conference proceeding or a Chapter in BookConference contribution

    1 Citation (Scopus)

    Abstract

    This paper presents a new approach based on a Two-Stage Linear Discriminant Analysis (Two-Stage LDA) and Conjugate Gradient Algorithms (CGAs) for face recognition. A Two-Stage LDA technique is proposed that utilises the null space of the sample covariance matrix as well as using the range space of the between-class scatter matrix to extract discriminant information. Classic Back Propagation (BP) is a widely used Neural Network (NN) training algorithm in many detectors and classifiers. However, it is both too slow for many practical problems and its performance is not satisfactory in many application areas, including face recognition. To overcome these problems, four CGA algorithms (Fletcher-Reeves CGA, Polak-Ribiere CGA, Powell-Beale CGA, scaled CGA) have been proposed, the utility of which we investigate here in combination with Two-Stage LDA features. To further improve the accuracy, a modified AdaBoost.M1 approach was employed, which combines results of several NN classifiers as a single strong classifier. Experiments are performed on the ORL, FERET and AR face databases. The results show that all of the proposed methods lead to increased recognition rates and shorter training times compared to the classic BP.
    Original languageEnglish
    Title of host publicationNeural Information Processing - LNCS 7667
    EditorsTingwen Huang, Zhigang Zeng, Chuangdong Li, Chi Sing Leung
    Place of PublicationBerlin Heidelberg
    PublisherSpringer
    Pages662-671
    Number of pages10
    ISBN (Print)9783642344992
    DOIs
    Publication statusPublished - 2012
    Event19th International Conference on Neural Information Processing 2012 - Doha, Doha, Qatar
    Duration: 12 Nov 201215 Nov 2012

    Conference

    Conference19th International Conference on Neural Information Processing 2012
    CountryQatar
    CityDoha
    Period12/11/1215/11/12

    Fingerprint

    Discriminant analysis
    Face recognition
    Neural networks
    Classifiers
    Backpropagation
    Adaptive boosting
    Covariance matrix
    Detectors

    Cite this

    Bozorgtabar, S., & Goecke, R. (2012). An Improved NN Training Scheme Using Two-Stage LDA Features for Face Recognition. In T. Huang, Z. Zeng, C. Li, & C. S. Leung (Eds.), Neural Information Processing - LNCS 7667 (pp. 662-671). Berlin Heidelberg: Springer. https://doi.org/10.1007/978-3-642-34500-5_78
    Bozorgtabar, Seyed ; Goecke, Roland. / An Improved NN Training Scheme Using Two-Stage LDA Features for Face Recognition. Neural Information Processing - LNCS 7667. editor / Tingwen Huang ; Zhigang Zeng ; Chuangdong Li ; Chi Sing Leung. Berlin Heidelberg : Springer, 2012. pp. 662-671
    @inproceedings{f3100363c37f47738809e777c20a8094,
    title = "An Improved NN Training Scheme Using Two-Stage LDA Features for Face Recognition",
    abstract = "This paper presents a new approach based on a Two-Stage Linear Discriminant Analysis (Two-Stage LDA) and Conjugate Gradient Algorithms (CGAs) for face recognition. A Two-Stage LDA technique is proposed that utilises the null space of the sample covariance matrix as well as using the range space of the between-class scatter matrix to extract discriminant information. Classic Back Propagation (BP) is a widely used Neural Network (NN) training algorithm in many detectors and classifiers. However, it is both too slow for many practical problems and its performance is not satisfactory in many application areas, including face recognition. To overcome these problems, four CGA algorithms (Fletcher-Reeves CGA, Polak-Ribiere CGA, Powell-Beale CGA, scaled CGA) have been proposed, the utility of which we investigate here in combination with Two-Stage LDA features. To further improve the accuracy, a modified AdaBoost.M1 approach was employed, which combines results of several NN classifiers as a single strong classifier. Experiments are performed on the ORL, FERET and AR face databases. The results show that all of the proposed methods lead to increased recognition rates and shorter training times compared to the classic BP.",
    keywords = "Neural Network, Face recognition, Linear Discriminant Analysis",
    author = "Seyed Bozorgtabar and Roland Goecke",
    year = "2012",
    doi = "10.1007/978-3-642-34500-5_78",
    language = "English",
    isbn = "9783642344992",
    pages = "662--671",
    editor = "Tingwen Huang and Zhigang Zeng and Chuangdong Li and Leung, {Chi Sing}",
    booktitle = "Neural Information Processing - LNCS 7667",
    publisher = "Springer",
    address = "Netherlands",

    }

    Bozorgtabar, S & Goecke, R 2012, An Improved NN Training Scheme Using Two-Stage LDA Features for Face Recognition. in T Huang, Z Zeng, C Li & CS Leung (eds), Neural Information Processing - LNCS 7667. Springer, Berlin Heidelberg, pp. 662-671, 19th International Conference on Neural Information Processing 2012, Doha, Qatar, 12/11/12. https://doi.org/10.1007/978-3-642-34500-5_78

    An Improved NN Training Scheme Using Two-Stage LDA Features for Face Recognition. / Bozorgtabar, Seyed; Goecke, Roland.

    Neural Information Processing - LNCS 7667. ed. / Tingwen Huang; Zhigang Zeng; Chuangdong Li; Chi Sing Leung. Berlin Heidelberg : Springer, 2012. p. 662-671.

    Research output: A Conference proceeding or a Chapter in BookConference contribution

    TY - GEN

    T1 - An Improved NN Training Scheme Using Two-Stage LDA Features for Face Recognition

    AU - Bozorgtabar, Seyed

    AU - Goecke, Roland

    PY - 2012

    Y1 - 2012

    N2 - This paper presents a new approach based on a Two-Stage Linear Discriminant Analysis (Two-Stage LDA) and Conjugate Gradient Algorithms (CGAs) for face recognition. A Two-Stage LDA technique is proposed that utilises the null space of the sample covariance matrix as well as using the range space of the between-class scatter matrix to extract discriminant information. Classic Back Propagation (BP) is a widely used Neural Network (NN) training algorithm in many detectors and classifiers. However, it is both too slow for many practical problems and its performance is not satisfactory in many application areas, including face recognition. To overcome these problems, four CGA algorithms (Fletcher-Reeves CGA, Polak-Ribiere CGA, Powell-Beale CGA, scaled CGA) have been proposed, the utility of which we investigate here in combination with Two-Stage LDA features. To further improve the accuracy, a modified AdaBoost.M1 approach was employed, which combines results of several NN classifiers as a single strong classifier. Experiments are performed on the ORL, FERET and AR face databases. The results show that all of the proposed methods lead to increased recognition rates and shorter training times compared to the classic BP.

    AB - This paper presents a new approach based on a Two-Stage Linear Discriminant Analysis (Two-Stage LDA) and Conjugate Gradient Algorithms (CGAs) for face recognition. A Two-Stage LDA technique is proposed that utilises the null space of the sample covariance matrix as well as using the range space of the between-class scatter matrix to extract discriminant information. Classic Back Propagation (BP) is a widely used Neural Network (NN) training algorithm in many detectors and classifiers. However, it is both too slow for many practical problems and its performance is not satisfactory in many application areas, including face recognition. To overcome these problems, four CGA algorithms (Fletcher-Reeves CGA, Polak-Ribiere CGA, Powell-Beale CGA, scaled CGA) have been proposed, the utility of which we investigate here in combination with Two-Stage LDA features. To further improve the accuracy, a modified AdaBoost.M1 approach was employed, which combines results of several NN classifiers as a single strong classifier. Experiments are performed on the ORL, FERET and AR face databases. The results show that all of the proposed methods lead to increased recognition rates and shorter training times compared to the classic BP.

    KW - Neural Network

    KW - Face recognition

    KW - Linear Discriminant Analysis

    U2 - 10.1007/978-3-642-34500-5_78

    DO - 10.1007/978-3-642-34500-5_78

    M3 - Conference contribution

    SN - 9783642344992

    SP - 662

    EP - 671

    BT - Neural Information Processing - LNCS 7667

    A2 - Huang, Tingwen

    A2 - Zeng, Zhigang

    A2 - Li, Chuangdong

    A2 - Leung, Chi Sing

    PB - Springer

    CY - Berlin Heidelberg

    ER -

    Bozorgtabar S, Goecke R. An Improved NN Training Scheme Using Two-Stage LDA Features for Face Recognition. In Huang T, Zeng Z, Li C, Leung CS, editors, Neural Information Processing - LNCS 7667. Berlin Heidelberg: Springer. 2012. p. 662-671 https://doi.org/10.1007/978-3-642-34500-5_78