About the Course
This is an online introductory course about the analysis and synthesis of digital signals. On completion you will be able to interpret and manipulate a variety of signals such as ECG (electrocardiograms – heart activity) and EEG (electroencephalogram – brain activity). While the course uses familiar signals (such as audio) to illustrate concepts, the content can be applied to a wide range of disciplines, such as telecommunications, energy systems and biomedical science.
Participants develop their signal processing skills by working on practical problems using Matlab or Octave after engaging with video tutorials, online quizzes and other resources.
- Feature Identification
- Frequency Domain Analysis of Signals
- Data Visualisation
- Discrete Fourier Transform
- Correlation Techniques
- Discrete Systems
- Filtering Signals
- Signal Synthesis
- From the Editor
- Recent Posts
- Popular (this month)
- Popular (all time)
- All Popular Tweets
- Vendors Only
- Free Books
- Free PDFs
Online DSP Classes: Why Such a High Dropout Rate?
Last year the IEEE Signal Processing Magazine published a lengthy
article describing three university-sponsored online digital signal
processing (DSP) courses . The article detailed all the effort the
professors expended in creating those courses and the courses’ perceived
values to students.
However, one fact that struck me as important, but not thoroughly addressed in the article, was the shocking dropout rate of those online courses. For two of the courses the article’s authors presented the following two tables. The tables’ rightmost columns, “Yield”, show the percentage of enrolled students who actually completed the courses.
[Tables reproduced from the IEEE article without permission.]
Shockingly, the average dropout rate is 97%.
I wondered, “Why do 97 out of every 100 online students drop out of the courses before completion?” My guess at the answer is that the courses are far too mathematically intense for the average young person who wants to learn something about DSP. On their home pages the EPFL course is described as “a mathematically solid introduction” to DSP. The Rice University course states that prospective students should have an “advanced mathematical background.”
Introductory DSP training material whose recipe is too rich in complex-variable mathematics is hard for beginners to digest. Most DSP beginners choke on such material and give up.
I believe another contributing factor to the online courses’ appalling completion rate is that these courses are not “self-paced”—-students must keep pace with the courses’ nonstop schedule. The Instructors’ predicted that the “effort” needed to complete the courses is 8 hours/week. My guess is that the average student ends up spending, say, 10 hours/week or more on a course. And that much intense effort, week after week, is more than many young students are willing to put forth.
Considering the courses’ dreadful completion rates, it seems to me the courses’ creators should now step back and ask “What can we do to improve our courses’ completion rates?”
[Postscript: Thinking more about this topic, I may be making a very false assumption that everyone who enrolled in an online DSP course actually intended to complete the course.]
References 5 are the home web pages of the three DSP courses discussed in the Reference  IEEE Sig. Proc. Magazine article. References 11 are other online DSP courses I discovered on the Internet.
 Thomas A. Baran, et al, “MOOC Adventures in Signal Processing”, IEEE Signal Processing Magazine, July 2016, pp. 62-83. https://infoscience.epfl.ch/record/221379/files/07503171.pdf
 École Polytechnique Fédérale de Lausanne’s “Digital Signal Processing” course on the Coursera.org web site.
 Rice University’s “Discrete-Time Signals and Systems, Part 1: Time Domain” course on the edX.org web site.
 Rice University’s “Discrete-Time Signals and Systems, Part 2: Frequency Domain” course on the edX.org web site.
 Massachusetts Institute of Technology’s “Discrete-Time Signal Processing” course on edX
 Universitat Pompeu Fabra of Barcelona and Stanford University’s “Audio Signal Processing for Music Applications” on the Coursera.org web site.
 Indian Institute of Technology Bombay‘s “Signals and Systems, Part 1”, on the edX.org web site.
 Indian Institute of Technology Bombay‘s “Signals and Systems, Part 2”, on the edX.org web site.
 “Fundamentals of Digital Image and Video Processing ” on the Coursera.org web site.
 A. V. Oppenheim, “1975 version of a videotape lecture series and study guide on digital signal processing,” MIT Center for Advanced Engineering Studies.
 A. V. Oppenheim. “1987 version of a videotape lecture series and study guide on digital signal processing,” MIT Center for Advanced Engineering Studies.
Previous post by Rick Lyons:
Errata for the book: ‘Understanding Digital Signal Processing’
Next post by Rick Lyons:
An Efficient Linear Interpolation Scheme
- Write a Comment
Select to add a comment
I think if you separate the online courses that are offered by accredited institutions such as Universities, require payment, hold students accountable, and have a clear prerequisite you will find the drop-out rate is lower (my intuition). Whereas a free (or low cost) on-line courses such as edx, coursera, MOOC, and others have a higher dropout rate. My rational is two-fold; first, a student that has a financial commitment to a course has an incentive to get their bang for their buck and may be perusing a degree, or is motivated enough to meet the financial commitment. Second, free on-line courses are often not maintained. I realize the phrase, “your mileage may vary” applies here as the free on-line course do have a bit of difference. Free on-line courses (especially coursera) do track student’s progress but there is no “penalty” to not complete, or, the certificates offered, IMHO, don’t really mean much. So, I think a key reason for lack free drop-outs could be that, like browsing for a book in a book store, one that “signs up” for the course could be simply testing the water to see what is “inside” and may decide the course is not for them ; complex variables or not.
All that said, books such as yours and other Authors on this site (and a few others) seem to have hit a sweet spot. Personally, I highly value both the print and those that take their time to publish on-line. I would like to recognize on site (I hope this does not break the rules on Embedded Related). That is Dr. Wickert at UCCS (www.eas.uccs.edu/wickert). Although Dr. Wikert does protect some material, he is generous enough to offer a lot of great material from his courses, and takes great pain to organize his material for his students, his primary goal is not to publish free materiel (That I can tell). He does offer the material for those that really want to learn on their own time, at a distance, without any offer to respond to email. One really nice feature of his classes/we site is lab demonstrations and configurations that have made a huge difference, in my opinion, in one can or should learn DSP – hands-on in conjunction with lectures. That’s not a new concept, but is evident in his “free” on-line material.
Finally, and I apologize for the verbose comment, I value HAM radio material and have thought about writing myself and approaching my writing similar to how you did in your books (disclaimer: I do not consider myself even 1/100000th the author you are. Your books are great). I have noticed that material, that I’ve read, seem seem to be shy about going into depth in the theory. I think there is a way to accomplish the simple, beginner hand-on, to the advanced with a bit of solid theory with hand-on, if a setup that just works, is presented with appropriate material. That’s one of my goals….
By the way, your recommendation on the ASN filter designer is one piece of advice I took. It’s really good. They have just added CMSIS support for the CORTEX M4. If only I can get PayPal tom complete my upgrade… 😉
I hope all is well in your part of the world.
As Always, I am A Rick Lyons Fan 🙂
Hi Dave. Nice to hear from you, and thanks for your thoughts. I agree that one of the big problems with online DSP courses is that you have to sign up for the class to see what topics are covered. (Sound familiar? “We have to pass the bill to see what’s in it.”)
I checked out Prof. Wickert’s web site. He has several courses that I’d like to attend. His “Real_Time Digital Signal Processing” class looks technically intense and awfully interesting, but being “C-language challenged” I’m not sure I’m up to such a task.
I am one of those dropouts, and I’d like to give my perspective.
I started the Stanford course on Coursera, link #6 above… I was pretty excited as audio is how I came to DSP in the first place. By day I am a software engineer. As of now I’ve been in that game 27 years now.. nothing fancy… but at about year 20 I got bored and thought well there has to be some domain I can apply my programming skills to, in order to do something not only different but *important*, you know, not “just a job” lol… At that time (oh, 2009) I thought Bioinformatics might be my calling, having read The Cosmic Serpent by Jeremy Narby a few years before that… After a year of Biology and an intro Bioinformatics course, I got burnt out and bored again.
Along came The Audio Programming Book from MIT Press, which I purchased in 2012. “My God,”, I thought, “why didn’t I think of that before?” … I bought it, got halfway into Chapter 1, and was flummoxed by the simple (well now it is for me anyway) math of exponential decay… So I was stuck there, and put it aside for what became a couple years, occasionally poking around trying to find something else to help me, but everything I found was even more difficult than that one…
At some point, perhaps in late 2015, along came some article on ChucK in one of my email newsletters, and my interest was rekindled. I’ve been a musician since I was a kid but never had any interest in computer music, until these two things came along… but now it’s to the point where I have, perhaps almost unintentionally, taken to immersing myself in DSP land in my free time, to try to absorb, think, and cross my fingers that someday I will get it before I get burnt out again. I think my chances are pretty good here since I am a musician as I mentioned. I fiddled around with ChucK for awhile and eventually completed the Kadenze course earlier this year.
Anyway, I am one of the dropouts in that table above. And yes, it was the math. When I started the course earlier this year, shortly after completing the Chuck course on Kadenze, I thought “here it is! my ticket outta here!” yet about 10 minutes into the first video lecture, there it appeared, the DFT formula, and thus, the wind immediately dissipated from my sails… Going in I knew that I was still short on the math, but I had mistakenly thought this was the one that was going to let me ride shotgun for awhile til I was ready to take the wheel. Nope.
I can’t speak for others, but for me at least, I DO plan on going back and making another run at that course after I feel like I’m ready. I am continuing my progress that I have made so far. I would say, things are actually starting to gel for me where I can read something, process it, and understand without having to necessarily go write the code to print out the mathematical results in order to grok it. I feel like I’m on the cusp of getting a good handle on the *basic* synthesis part of things (sine/square/triangle/sawtooth/pulse/chirp), at least enough to understand various waveforms, various ways to synthesize them, and whatnot. I know how to read a basic spectrogram… just not how to compute it yet. Thank you to Mr. Boulanger (I’m currently hammering around on chapters 2/3/4 in TAPB) and to Mr. Kapur and Mr. Cook (finished the course, and am looking forward to trying out the synthesis tools).
I’m rambling, but I wanted to give you the complete perspective of how I ended up here. I aced math in high school. I mean I had the rhythm and everything — it was the one subject I knew I could ace without having to really think too hard. It just came naturally to me. So did the drums, as an aside. Fast forward 30 years (I turned 48 last week on 10.17), and I find myself wishing I would have either A) perhaps finished a college degree, or at least the Calculus part, or B) discovered my interest in computer music sooner! I said I love math, and I still do. It doesn’t come *as* easy as it did in high school, but if I sit and work with it, I’m still “pretty good” at it — I just don’t have the rhythm down anymore since I’m out of practice.
Earlier in the year I finished my own review of Basic Algebra with a text I bought myself, and am currently getting started on a Pre-Calculus Algebra book. The main thing that bothers me now is that I’m impatient. I can assume that there’s a good chance that I will live long enough to master this subject… but what drags me down is that I know that if I work my way through this PCA text, I may end up using only pieces of it in the end, and in the end that’s what matters — *using* the math, to synthesize and process audio, because I’m going to forget the rest of the math that I won’t use in that field, so of course when I’m doing the exercises for whatever chapter, part of me is thinking “am I wasting my time by doing these exercises?”. Perhaps my viewpoint is a bit off there and I’m sure someone will correct me if so. I recently came across the website here and the Mathematics of the DFT book may end up being what saves me and propels me back into that Coursera course and across the finish line…. at least it has given me that hope back that I won’t have to spend too much time on the math that I won’t end up using 😉
Hello sidtbird. Thanks for sharing your story. Just what mathematics topics are useful to study in preparation for practicing digital signal processing is a source of endless debate.
Sidtbird, would you please send me a private e-mail at: <[email protected]>?
This makes me wonder, how many people who visit our blog entries on DSPRelated find them helpful, or for that matter read them in whole or in part.
I recall reading an article a few years ago on Massive Online Open Courses (MOOCs) in engineering — apparently the initial offerings had horrible completion rates. There had been a lot of hype on how they would change education.
Maybe just sitting in a room with other humans is underrated.
Question: Did the students pay for the courses, or were they free?
Hi Neil. You wrote: “This makes me wonder, how many people
who visit our blog entries on DSPRelated find them helpful, or for that
matter read them in whole or in part.” I’d sure like to know that answer to that. I’ll bet that less than 20% of the folks who start reading our blogs end up reading them in full.
As for your question, as far as I can tell the three online DSP courses are free but the EPFL course offers some sort of official “Certificate” if the student pays $40 and competes the course.
Neil, thinking more about this topic, I may be making a very false assumption that everyone who enrolled in an online DSP course actually intended to complete the course.
In my experience if you give something away for free, even if it’s extremely valuable, people are happy to disregard its value. When people pay for a thing they are far more likely to see it through as they have investment to recoup.
Hi andrewstanfordjason. “Bingo! Give that young man a kewpie doll.”
I wish our State and Federal politicians understood human nature as well as you do.
Where’s your control group? What’s the completion rate for courses with similar estimated levels of difficulty and esotericity (I just made that word up) in radically different fields — i.e., statistical methods in early childhood education research, 2nd-kingdom Egyptology, genetic factors in oncology, etc.
It’s possible that a large percentage of people sign up for the course, click into a few videos, feel uninterested, and hence drop. I wonder what is P(drops the course eventually | completed the first half lectures & exercises) ?
Hello khuasw. I don’t know the answer to your question. However, perhaps the answer is buried somewhere in the article given in my above Reference .
Thank you Rick.
“those who survive Fourier seem to keep
their interest alive throughout the rest of the course”  pg. 66
This is actually not a completely bad result. The same thing happens in college courses while the online audience generally takes a less serious approach and it can be expected that they’re on a lower “weed out” threshold.
The biggest issue with these courses is that they are not ‘self paced’. You are forced to keep up with the class. I am assuming the majority of the people that are taking these courses are either A) full-time students, or B) have a full-time job.
In both these scenarios, its really challenging to find time to commit yourself. If I am going to be “self-studying”, I need to be able to go at my own pace.
Let’s say I come across a topic I dont understand fully, I’ll probably want to explore other resources and take my time to really make sure I have a good grasp of this before I move forward. Now if I was in a traditional academic structure / model, I would be bombarded with assignments, labs, projects, even though I lack understand of the subject material being discussed. The professor would not wait for me, and the class would keep moving. That gap in my knowledge will accumulate over time causing further lack of understanding. These online courses are not much different.
I was really looking forward to taking some of the courses mentioned here, but unfortunately, I simply dont have the time at the moment. Truly upsetting, because I am sure, content wise, these courses would be really beneficial.
Rick, responding to your statement “I believe another contributing factor to the online courses’ appalling completion rate is that these courses are not “self-paced”” I can see that this would be a problem. I am almost finished with a Udacity class on Artificial Intelligence Deep Learning. The class is self paced within limits. Each of the five projects has a due date but if you miss the date by a day or two, it’s not a big deal. The downside of lateness is that you fall behind your fellow students who share Q/A on the forum. I think Udacity also rewards on-time submissions with an occasional freeby for another class.
To post reply to a comment, click on the ‘reply’ button attached to each comment. To post a new comment (not a reply to a comment) check out the ‘Write a Comment’ tab at the top of the comments.
Registering will allow you to participate to the forums on ALL the related sites and give you access to all pdf downloads.
EE Job Opportunities
Communications/DSP Engineer (San Francisco, CA)
Work with a small team of signal processing/communications engineers to develop and implement various radio functions (filtering, gain control, frequency tracking, digital upconversion/downconversion, linearization, timing and phase recovery, equalization, FEC, symbol mapping/demapping, framing, etc )
You might also like…
Subscribe to this Blog
Receive a notification when Rick Lyons publishes a new article:
About Rick Lyons
Richard Lyons is a Contracting Systems Engineer and Lecturer at Besser Associates, Mountain View, Calif. He has written over 30 articles and conference papers on DSP topics, and authored Amazon.com’s top selling DSP book ” Understanding Digital Signal Processing, 3rd Ed. “. He served as an Associate Editor at IEEE Signal Processing Magazine, for nine years, where he created and edited the “DSP Tips & Tricks” column. Lyons is the editor of, and contributor to, the book ” Streamlining Digital Signal Processing-A Tricks of the Trade Guidebook, 2nd Ed. ” (Wiley & Sons, 2012).
Books by Rick Lyons
Popular Posts by Rick Lyons
- A Quadrature Signals Tutorial: Complex, But Not Complicated
- Digital Envelope Detection: The Good, the Bad, and the Ugly
- Understanding the ‘Phasing Method’ of Single Sideband Demodulation
- Four Ways to Compute an Inverse FFT Using the Forward FFT Algorithm
- Free DSP Books on the Internet
Blogs – Hall of Fame
A Fixed-Point Introduction by Example
Handling Spectral Inversion in Baseband Processing
Understanding the Phasing Method of Single Sideband Modulation
An Interesting Fourier Transform 1/f Noise
Free PDF Downloads
FFT Interpolation Based on FFT Samples: A Detective Story With a Surprise Ending
Adaptive Algorithms in Digital Signal Processing – Overview, Theory and Applications
Algorithms for Efficient Computation of Convolution
- Free Books
- Free PDFs
- Terms of Service
- Cookies Policy