After all these implements and texts designed by intellects we're vexed to find evidently there's still so much that hides.
May 15, 2011 8:47 AM   Subscribe

I like math! I took an introductory undergraduate math course on DSP and, well, I want more.

This was an upper-level undergraduate course — most of the students were graduating seniors. I myself am a few years out of undergrad and work in a cognitive neuroscience lab. Understanding DSP would be quite useful to me! My engineering background is nil and my math background is what I've managed to piece together through courses taken in my spare time.

DSP seems to be a vast subject. The class barely scratched the surface. We introduced Fourier series, the Fourier transform, DFT (as well as the DHT and DCT), the FFT algorithm, and looked briefly at wavelets. But all this was only covered in superficial detail, because it was still too much and most of the students in the class were adrift. Based on my final grade, I was apparently less adrift than most, but I still feel that I barely understand the material we did cover.

So, my questions:
  1. How is DSP typically covered in math and EE departments? The Internet suggests that undergrads typically only get a class or two on DSP, and that it's mostly studied at the graduate level. Is this accurate?
  2. If I did decide to take a graduate course on DSP, what sort of background would I need?
  3. I know nothing about circuits, processors, or "physical stuff." How far can I get into DSP without some knowledge of analog things?
  4. Searching Amazon suggests that Oppenheim is the DSP bible. Lyons is suggested as a gentle introduction, and Proakis is somewhere between them. I am not math-averse, and I'd prefer to avoid buying multiple expensive textbooks. I'm leaning toward the Proakis, which sounds like a good learning resource, but also a good reference. Alternative suggestions?
I appreciate your advice.
posted by Nomyte to Education (2 answers total) 8 users marked this as a favorite
1) At my university, there was no Math department coverage of DSP. There were courses in algorithm development and logic, but these were oriented towards the computer sciences rather than engineering. For Electrical Engineering, students were required to take a semester in a class that was about discrete systems, and would have roughly been at the same level as the course you took. However, it was oriented more towards a traditional "control systems" perspective than a DSP perspective.

2) This strongly depends on the university where you study. DSP as a field of study is essentially applied mathematics and requires very little outside material. However, not all universities teach pure DSP. For instance, wireless communications may be taught, incorporating DSP, but also incorporating RF principles which would require an background in electromagnetics. For what it's worth, once you get past the introductory level stuff, there is less correlation to "the real world" than you might expect. For instance, there is very little "physical" correlation to a FIR filter, although there is a strong "physical" correlation to a IIR filter.

3) Surprisingly far. I worked with two different DSP engineers whose background is not primarily in electrical engineering. Both took a non-EE background (one in Physics, the other in Math), and then did a MS in EE and Communications Systems, respectively.

4) The Lyons book is useful, but very applied. I recommend it, but if you want to buy only one textbook, I suggest Proakis (which I do not own, but have used). I own Proakis' Digital Communications, which is as complete a reference that I know about - and he definitely doesn't shy away from the underpinnings of the subject.
posted by saeculorum at 9:13 AM on May 15, 2011

2. Look up the grad classes you want to take, they should list prerequisites. Alternately, speak with the professors.

3. I'd be surprised if you need to know any "physical" stuff. Unless you are building some very exotic special-purpose hardware, you should be able to do plenty of DSP work in software. The Intel architecture has included DSP friendly vector operations for several generations now, so you should be able to do a decent amount of DSP work on an ordinary desktop / laptop.

4. Again, probably best to look up the professors whose classes you will take.

FWIW - I'm a software guy with an interest in audio & finance DSP. But I haven't studied either much.
posted by b1tr0t at 10:45 AM on May 15, 2011

« Older Good Short Fiction Set in Kentucky?   |   How do I control my rage? Newer »
This thread is closed to new comments.