Human information retrieval by face extraction and recognition on TV news images using subspace method
This paper proposes a human information retrieval system with face extraction and recognition function. A user watches the TV and if he encounters with interesting but unknown persons, he can specify them on the TV display by a mouse device. The proposed system can extract and recognize the specified persons and retrieve their information from human information database via interet and output it on a display unit as well as through voice synthesizer. We used subspace methods for a face extraction and recognition. In face extraction, the sub-image within a scanning window is projected to facial subspace and it is regarded as the facial region, if the projection amount has a local peak. In face recognition, the facial subspace of individual person is constructed at first in an observation space using respective training data. Then an input facial region is projected to the individual subspace and classified into the person with the maximum projection amount.
Unable to display preview. Download preview PDF.
- 1.Y.Ariki, N.Ishikawa: Integration of Face and Speaker Recognition by Subspace method. ICPR96. (1996) C456–460Google Scholar
- 2.Y.Sugiyama, Y.Ariki: Facial region tracking and Recognition by Subspace Projection. VSMM96. (1996) 225–230Google Scholar
- 3.Y.Ariki, N.Ishikawa and Y.sugiyama: Extraction and Recognition of Facial Regions by Subspace Method. ACCV95. (1995) III-733–III-742Google Scholar
- 4.E.Oja: Subspace Methods of Pattern Recognition. Research Studies Press. England. (1983)Google Scholar
- 5.M.A.Turk and A.P.Pentland: Face Recognition using Eigenfaces Proc. of IEEE Conference on Computer Vision and Pattern Recognition. (1991) 586–591Google Scholar