部分空間を用いたパターン認識
研究概要
文字や顔等、この世界に存在するさまざまな物体を圧縮して表現することで、それらを高速に高い精度で機械に認識させる方法について研究を行っています。物体（パターン）を圧縮することで、認識に重要な成分を抽出することができ、さらに少ない記憶容量でさまざまなパターンを表現することが可能となることから、高速化や高精度化が容易となります。この性質を利用すれば、計算能力が低く、少ない記憶容量しか持たない携帯端末上で物体を認識・検出したり、機械にとっては難しい認識問題を効率よく解決できたりします。
主要論文・参考事項
Y. Washizawa and S. HOTTA, ''Mahalanobis Distance on Extended Grassmann Manifolds for Variational Pattern Analysis,'' IEEE Trans. on Neural Networks and Learning Systems, accepted, 2014. | doi: 10.1109/TNNLS.2014.2301178
Hiroto MINEGISHI and Seiji HOTTA, ''Vote-Based Image Classification Using Linear Manifolds,'' J. of IIEEJ, vol.40, no.1, pp.52-58, 2011.
Yosuke SHIMIZU and Seiji HOTTA, ''Detection and Retrieval of Nucleated Red Blood Cells Using Linear Subspaces,'' J. of IIEEJ, vol.40, no.1, pp.67-73, 2011.
Kzauki KONDO and Seiji HOTTA, ''Color Image Classification Using Block Matching and Learning,'' IEICE Trans. on Info. & Sys., vol.E92-D, no.7, pp.1484-1487, 2009.
Seiji HOTTA, ''Generalized Learning Local Averaging Classifier,'' J. of IIEEJ, vol.37, no.3, pp.206-213, 2008 (Paper Award from IIEEJ in 2010).
Seiji HOTTA, ''Local Subspace Classifier with Transform-Invariance for Image Classification,'' IEICE Trans. on Info. & Sys., vol.E91-D, no.6, pp.1756-1763, 2008.
Seiji HOTTA, Senya KIYASU, and Sueharu MIYAHARA, ''Arbitrary-Shaped Cluster Separation Using One-Dimensional Data Mapping and Histogram Segmentation,'' Journal of Advanced Computational Intelligence and Intelligent Informatics, Vol.11, No.9, pp.1136-1143, 2007.
お問い合わせ先
東京農工大学・先端産学連携研究推進センター
urac[at]ml.tuat.ac.jp（[at]を@に変換してください）
Pattern Recognition by Using Subspace Method
Research members： Dr. Seiji Hotta
Research fields： Principles of Informatics
Departments： Institute of Engineering
Keywords： pattern recognition
Web site：
Summary
Classifier design is one of the most important issues in the pattern recognition domain. Our laboratory focuses on a linear classifier called subspace method. This method maps high-dimensional samples into a low-dimensional subspace obtained by eigenvalue decomposition of an autocorrelation matrix. By using this mapping, a high-dimensional sample can be represented by a linear combination of a small number of eigenvectors. This property yields high accuracy and speed in a classification phase.
Reference articles and patents
Y. Washizawa and S. HOTTA, ''Mahalanobis Distance on Extended Grassmann Manifolds for Variational Pattern Analysis,'' IEEE Trans. on Neural Networks and Learning Systems, accepted, 2014. | doi: 10.1109/TNNLS.2014.2301178
Hiroto MINEGISHI and Seiji HOTTA, ''Vote-Based Image Classification Using Linear Manifolds,'' J. of IIEEJ, vol.40, no.1, pp.52-58, 2011.
Yosuke SHIMIZU and Seiji HOTTA, ''Detection and Retrieval of Nucleated Red Blood Cells Using Linear Subspaces,'' J. of IIEEJ, vol.40, no.1, pp.67-73, 2011.
Kzauki KONDO and Seiji HOTTA, ''Color Image Classification Using Block Matching and Learning,'' IEICE Trans. on Info. & Sys., vol.E92-D, no.7, pp.1484-1487, 2009.
Seiji HOTTA, ''Generalized Learning Local Averaging Classifier,'' J. of IIEEJ, vol.37, no.3, pp.206-213, 2008 (Paper Award from IIEEJ in 2010).
Seiji HOTTA, ''Local Subspace Classifier with Transform-Invariance for Image Classification,'' IEICE Trans. on Info. & Sys., vol.E91-D, no.6, pp.1756-1763, 2008.
Seiji HOTTA, Senya KIYASU, and Sueharu MIYAHARA, ''Arbitrary-Shaped Cluster Separation Using One-Dimensional Data Mapping and Histogram Segmentation,'' Journal of Advanced Computational Intelligence and Intelligent Informatics, Vol.11, No.9, pp.1136-1143, 2007.
Contact
University Research Administration Center(URAC),
Tokyo University of Agriculture andTechnology
urac[at]ml.tuat.ac.jp
(Please replace [at] with @.)