Finally I would like to thank the word processing program LATEX and the mathe- ... Chapter 2 describes the concept of canonical correlation. This you have to know .... where A is a linear transformation matrix, t is a time-shift sometimes called.
Multidimensional signal recognition, invariant to ane transformation and time-shift, using canonical correlation Thesis project done at Computer Vision Laboratory Linkoping University Sweden by
Bjorn Johansson
Reg. nr: LiTH-ISY-EX-1825
Supervisor: Magnus Borga Examiner: Hans Knutsson Linkoping, October 6, 1997
Introduction 1.1 Acknowledgements I would like to thank my supervisor licentiate Magnus Borga and my examiner associate professor Hans Knutsson for many inspiring discussions and ideas that has moved the work forward (I got no help from literature - could not nd any concerning this topic). I would also like to thank Magnus Borga for all constructive criticism in proof-reading my thesis. Thank you also to the rest of the sta at the Computer Vision laboratory for helping me with practical things during the work. Thanks also to Stefan Eriksson for taking on the job as opponent. Finally I would like to thank the word processing program LATEX and the mathematical software MATLAB for being cooperative (most of the time).
1.2 Notation Italics (x) are used for scalars. Lowercase letters in boldface (x) are used for vectors. Uppercase letters in boldface (X) are used for matrices. Transpose is denoted by a 'T ' (e.g. xT ). Adjungation (transpose + conjugation) is denoted by a '' (e.g. x ). '^' over a vector indicates unit length (e.g. jv^ j = 1.)