Next:
Introduction
Up:
A System for Recognising
Previous:
A System for Recognising
Contents
Introduction
Discussion of Related Work
Eigenlips
2D Optical Flow
Deformable Templates & Active Contours
Relevant Processing Techniques
Analysis of Local Orientation and Phase
Orientation Template Correlation for Feature Localisation
Representation and Visualisation of Local Orientation Estimates
Methods for Estimating Local Orientation and Phase
Tensor Field Filtering
Active Contour Models
Optical Flow
Differential Methods
The Onedimensional Case
Principle Components Analysis
Gaze Estimation
Motivation
Pinhole Cameras and Perspective Projection
Estimation of the Perspective Projection Matrix
Classification
The Problem
An Optimal Decision Rule
The Probabilistic Eigenspace Approach
Discussion of the Proposed System
Localisation of the Mouth
Initialisation of the Feature Templates
Tracking the Mouth
Tracking the Vertical Lip Motion
Estimation of Mouth Elongation
Word Recognition
Experimental Results
Speaker Dependent Recognition
Speaker Independent Recognition
Possible Extensions and Suggestions for Future Work
Conclusion
Acknowledgements
Mathematical Supplement
References
About this document ...
Markus Weber
Tue Jan 7 15:44:13 PST 1997