I am doing b.e electrical and electronics from bits-pilani , india.

I have just completed my second year.

I had a course on signal and systems(fourier, laplace, sampling etc.) in my fourth semester.It has aroused my interest in dsp field.

Can anyone here guide me on how to proceed with my education in this field so that i could have both theoritical and practical knowledge in the field.

Also is it necessary to choose a specific field such as image, audio , video processing right from the beginning or should i take a course on general dsp first?

And my last question is directed towards indian members on this forum.I know of the huge scope of dsp outside. Does India has the same scope for a career in dsp field..

Hi,

I'm at a professor at the University of Calgary, Canada. I do research into both DSP and embedded systems.

I teach a 3rd year course on Assembly language and interfacing (using the Blackfin) -- essentially straight embedded systems. http://enel.ucalgary.ca/People/Smith/2008webs/encm511_08/

and a 4th year course of DSP processor architectures where I look at how the characteristic of DSP processors match the characateristics of DSP algorithms using the TigerSHARC http://enel.ucalgary.ca/People/Smith/2008webs/encm515_08/

Here is my take of the situation

First you must recognize the difference between embedded systems and DSP.

Most embedded systems on the market don't do DSP as they are applied in an area where DSP type calculations are not needed

I have many graduate students -- biomedical engineers -- who do DSP in the biomedical area. They are not really interested in "embedded systems" . They work in Matlab doing DSP on various images and biomedical signals. How do you improve this algorithm to improve diagnosis -- is the question they are trying to answer. Most of those algorithm remain "off" DSP processors and in Matlab. The algorithms remain off an embedded system unless the students decide to commercialize the DSP, and even then most of the students would only move the algorithms onto the existing systems that generate the images in the first place -- MRI scanners for example. A problem here is that it takes considerable effort to move Matlab code in an efficient manner onto an embedded system for commercialization-- more time than these students have inside a M. Sc. program.

I also have other students -- software engineers -- who are looking at the development of testing frameworks for designing reliable embedded systems. They are not really interested in DSP, and many have difficultly in moving their testing ideas from the normally taught desk-top environment over to embedded systems. However, their testing frameworks have to have an application to test, so I often get them to move one of the DSP algorithms from the biomedical enginering students onto an embedded system and show they can validate it.

Finally I have students "computer engineering" students who are interested in making DSP algorithms work on embedded systems. Many of these student do work in telecommunications where they use DSP algorithms to characterize the telecommunications channels e.g. to compensate for the peaks and valleys in the transmission characteristics and improve performance.

In previous years most of my students went into telecommunications area as the DSP algorithms from the biomedical area are exactly the same ones for telecommunications and are the same as the ones in XXXX etc., just applied in different areas.

For some reason, the last four of my students have all gone into industry to do DSP applications running on embedded systems for startup companies. This surprized me as I would not have seen Calgary as a hot bed for embedded developers. However perhaps the number of students I have taught about embedded systems over the last 20 years and remaining in Calgary has reached a critical mass :-)

To answer your question -- what you should do

Y

ou will need general DSP courses (Linear systems etc) and seperate embedded system courses. You will probably not find that they mix until 4th year where you might find specific technical electives. Personally, I find that image processing courses (independent of embedded systems) the most interesting as you can see what is happening when you apply the algorithms.If you do a M. Sc. -- then embedded systems are difficult -- they are applications rather than new ideas so difficult to get published. However, there are many biomedical and telecommunications situations where the embedded system component is important in the field studies. However you will get more credit for the field studies and not the embedded systems.

There is also much interest in system of a chip ideas.

Hope some of these ideas help

Message was edited by: Mike Smith