First of all, I would like to thank Evolution for coming up with the brilliant
idea of using the same device for chewing, breathing, and emitting sounds
for communication, making automatic speech recognition one of the most
challenging engineering tasks one can think of.
I am grateful to my colleagues for making my work environment so stimulating,
both at KTH, and during my stay as a guest researcher at ATR. In particular
Rolf Carlson has given important guidance over the whole period. It was
also studies of Rolf Carlson and James Glass that inspired to my work in
the area of speaker adaptation. Michael Phillips deserves acknowledgment
for introducing me to the A* algorithm and lexical search algorithms, and
Kjell Elenius has always been a good conversation partner in ANN related
Further, I would like to thank Björn Granström for allowing me
to work in this area, and for guidance and inspiration. I would also like
to thank Yoshinori Sagisaka for inviting me to work in the stimulating
research environment at ATR in Kyoto, Japan.
I also take the opportunity to thank all my teachers, all the way from
kindergarten to graduate course teachers. Of course, this thesis would
have been impossible without them.
I thank my mother and father for all their support, and for providing a
home where education was a natural part of life. All my friends deserve
acknowledgment for being so patient with me when I occasionally completely
forget about everything but my work. Finally, thank you Linda for love,
support, and understanding.
My first two years of doctoral studies were sponsored by a donation
by VOLVO AB, and the remaining studies were financially supported by NUTEK
and HSFR, and during the last year also by CTT, a center for speech technology,
jointly sponsored by NUTEK, Swedish industry, and KTH.