|
When I was being interviewed at the handwriting recognition group of IBM T.J.
Watson Research Center in December of 1990, one of the interviewers asked me
why, being a mechanical engineer, I was applying for a position in that group. Well,
he was an electrical engineer and somehow was under the impression that handwriting
recognition was an electrical engineering field! My response was that I had
done research on Kinematics, Dynamics, Control, Signal Processing, Optimization,
Neural Network Learning theory and lossless image compression during the past
7 years while I was in graduate school. I asked him what background he thought
would have been more relevant to do research in handwriting recognition.
Anyhow, I joined the on-line handwriting recognition group which worked sideby-
side with the speech recognition group. Later, I transferred to the speech recognition
group and worked on speaker recognition. Aside from the immediate front-end
processing, on-line handwriting recognition, signature verification, speech recognition
and speaker recognition have a lot in common. During the 10 years at IBM I
also worked on many complementary problems such as phonetics, statistical learning
theory, language modeling, information theoretic research, etc. This continued
with further work on real-time large-scale optimization, interactive voice response
systems, standardization and more detailed speaker recognition research at Recognition
Technologies, Inc. to the present date, not to mention the many years of code
optimization, integer arithmetic, software architecture and alike within the past 25
years. |