Feature space mutual information in speech-video sequences

T. Butz1, J.-P. Thiran1
1Signal Processing Institute (ITS), Swiss Federal Institute of Technology, Lausanne, Switzerland

Tóm tắt

We present an approach to directly study mutual relationships between audio and video signals for multimedia applications. The presented approach is mathematically based on information theory and is closely related to information theoretic classification. We show that very simple features of the audio- resp. video-channel can already contain lots of mutual information between both modalities. The mathematical approach is very general though and not restricted to the presented multimedia application.

Từ khóa

#Mutual information #Mouth #Space technology #Video signal processing #Feature extraction #Data mining #World Wide Web #Marine vehicles #Signal processing #Information theory

Tài liệu tham khảo

butz, 2002, Multi-modal signal processing: An information theoretical framework, Tech Rep 02 01 Signal Processing Institute (ITS) Swiss Federal Institute of Technology (EPFL) cutler, 2000, Look who's talking: Speaker detection using video and audio correlation, IEEE International Conference on Multimedia and Expo New York USA, 10.1109/ICME.2000.871073 cover, 1991, Elements of Information Theory, 10.1002/0471200611 fisher, 2000, Learning joint statistical models for audio-visual fusion and segregation, Advances in Neural Informotion Processing Systems Denver USA 10.1109/42.563664 10.1016/S1361-8415(01)80004-9 furui, 1994, An overview of speaker recognition technology, Proc ESCA Workshop on Automatic Speaker Recognition Identification and Verification, 1 10.1109/CVPR.2000.854730 fano, 1961, Transmission of Information A Statistical Theory of Communication devroye, 1985, Non-Parametric Density Estimation 10.1103/PhysRev.106.620 principe, 2000, Learning from examples with information theoretic criteria, Multimedia Signal Processing