| < Note 699.0 by MIZPAH::ELKIND "Steve Elkind" >
-< Any DSP course or mavens? >-
A colleague of mine needs to get up to speed on digital signal processing,
especially in regard to the application of vector machines to doing same.
Could anyone out there who:
1. Regards her(him)self as knowledgeable in this area
-and/or-
2. Is aware of internal or external courses in DSP (short, ~ 1-2 weeks
or less in length) ^
> The Northeastern State of the Art program offers several
> 10 week (one 2.5 hour night per week) courses in this area.
> I took the first course entitled (strangely enough)
> 'Digital Signal Processing' and found the instructor to be
> SUPERB (his name was John O'Donnell; he works at Zoran, or did
> until recently). He is INTIMATELY familiar with
> both the theory and practice of DSP. Just taking a course under someone
> this knowledgeable was a delightful experience (this is in contrast
> to the generally abysmal quality of the technical training available
> to the night student in the Massachusetts area). He hands out design
> programs as part of the class and explains things very well.
> I should mention that the focus is on application of the
> theory to design and specific design techniques for digital filters;
> for example, FIR design techniques, IIR elliptic filters.
> There is only peripheral discussion of things like power spectrum
> analysis.
-and/or-
3. Can suggest a book which might be useful for self-help?
> I'm not aware of any good books in this area that aren't either
> difficult to read on your own (ex. Rabiner and Gold)
> or too simple or inaccurate to bother with.
> The ones written by mathematicians are generally abstruse,
> and leave you wondering how you would really apply the stuff.
> The ones written by engineers don't give a unified treatment
> and the authors misconstrue the significance of some of the
> material.
One thing I should mention is that his math background is most likely at
the high school level, with whatever else the DoD taught him in technician
^
> Uh oh. This is a bit of a problem. Anyone not fluent with
> complex variable level math, a smattering of convolution, and some
> analog filter basics is going to have tough hoeing in the Northeastern
> sequence and the literature.
> I still recommend one of the Northeastern courses
> (make sure you get O'Donnell as the instructor). If he asks a lot
> of questions, he'll get good answers. Also there are others that
> Northeastern offers that I have not taken yet. If O'Donnell teaches
> any of them, they will be good.
training. He will not be writing code or anything, but he does need to
understand the major technical/technology issues at some level.
Please send mail to me at CSSE::ELKIND or call me at dtn 276-9100
Thanks much.
P.S. This request has also been posted in the HPSC (High-Perf. Sci. Comp)
notes file.
|
| I can suggest one introductory book - R. Hammings 'Digital Filters',
(second edition). This is one of the best introductory books and
gives the bare minimum of material one must know about. Hamming,
of course, is one of the old masters from Bell Labs in the areas
of digital filters, numerical analysis, and coding theory: the
Hamming code and Hamming windows are named after him.
For an introduction to FFT's that is very easy to read, look at
E. O. Brighams little book 'The Fast Fourier Transform'.
The subject of DSP is so concrete, that you really don't need an
extreme mathematical background to gain entry. At the same
time, the subject is extemely pervasive, and touches on areas like
signal analysis, numerical analysis, number theory, complex analysis,
filter synthesis, and on to the applications like image processing.
To have a deep understanding of it, you need a good math background
of course, but because of the broadness of digital signal processing
someone who has worked in a technical area will undoubtedly have
intuitive background in related subjects, and can get going quickly.
And many of the more advanced topics are only important to someone
actually implementing state of the art algorithms.
Offhand, I don't know of a reference on array processors per se, though
many papers have been written on that subject (but these are probably
not tutorial in nature).
The course mentioned in .1 may be a good way to go if adequate time
is available. No one textbook can really do DSP justice, so it is
very helpful to have an instructor who has read widely on the subject.
Hope this helps...
- Jim
|
|
There is a small (200 page) book entitled "Signal Processor Chips"
David Quarmby, editor, Prentice-Hall which gives a fairly elementary
overview of DSP filter algorithms and then has a chapter on each of three
or four current DSP chips: the Intel, NEC, TMS & ??. These are written by
in house people from each mfr. Not really array processing DSP but it's
a start...
I think the SoftPro bookstore in Burlington MA is where I saw it.
|