Just a reminder that if you are getting some value out of the CSP Blog, I would appreciate it if you could make a donation to offset my costs: I do pay WordPress to keep ads off the site! I also pay extra for a class of service that allows me to post large data sets like the one for the Machine-Learner Challenge.
If everyone that derived value from the CSP Blog were to donate $5, I’d have enough leftover for at least a couple cups of fancy coffee.
Where does IQ (or I/Q) data come from?
Previous SPTK Post: Digital Filters Next SPTK Post: TBD
Let’s really get into the mathematical details of “IQ data,” a phrase that appears in many CSP Blog posts and an awful lot of machine-learning papers on modulation recognition. Just what are “I” and “Q” anyway?
Continue reading “SPTK: I and Q”“Alive in the Superunknown
First it steals your Mind, and then it steals your … Soul”
–Soundgarden
An advantage of using and understanding the statistics of communication signals ™, the basics of signal processing, and the rich details of cyclostationary signal processing is that a practitioner can deal with, to some useful degree, unknown unknowns. The unknown unknowns I’m talking about here on the CSP Blog are, of course, signals. We know about the by-now-familiar known-type detection, multi-class modulation-recognition, and RF scene-analysis problems, in which it is often assumed that we know the signals we are looking for, but we don’t know their times of arrival, some of their parameters, or how they might overlap in time, frequency, and space. Then there are the less-familiar problems involving unknown unknowns.
Sometimes we just don’t know the signals we are looking for. We still want to do as good a job on RF scene analysis as we can, but there might be signals in the scene that do not conform to the body of knowledge we have, to date, of manmade RF signals. Or, in modern parlance, we didn’t even know we left such signals out of our neural-network training dataset; we’re a couple steps back from even worrying about generalization, because we don’t even know we can’t generalize since we are ignorant about what to generalize to.
In this post I look at the broadcast TV band, seen in downtown Monterey, California, sometime in the recent past. I expect to see ATSC DTV signals (of the older 8VSB/16VSB or the newer OFDM types), and I do. But what else is there? Spoiler: Unknown unknowns.
Let’s take a look.
Continue reading “Desultory CSP: What’s That Under the TV?”Another dataset aimed at the continuing problem of generalization in machine-learning-based modulation recognition. This one is a companion to CSPB.ML.2023, which features cochannel situations.
Quality datasets containing digital signals with varied parameters and lengths sufficient to permit many kinds of validation checks by signal-processing experts remain in short supply. In this post, we continue our efforts to provide such datasets by offering a companion unlabeled dataset to CSPB.ML.2023.
Continue reading “CSPB.ML.2023G1”A kvetch.
As the generative-AI crowd continues to feast on copyrighted material of all kinds, they are getting pushback in the form of lawsuits from artists, writers, and journalists. I discussed this recently with Dan and Eunice on the CSP Blog.
Open AI in particular seems to believe they have some kind of divine right to pursue whatever business they want, whether it is legal or not. Because reasons … including national security … and “meeting the needs of today’s citizens.” But probably just greed and hubris.
In a statement to the UK’s House of Lords, Open AI says this, and I assume they did so with a straight face, which would have been admirably difficult:
Continue reading “Stupid Laws Getting In My Way”A look at general linear time-invariant filtering in the discrete-time domain.
Previous SPTK Post: The Z Transform   Next SPTK Post: IQ Data
Linear shift-invariant systems are often called digital filters when they are designed objects as opposed to found objects, which are models, really, of systems occurring in the natural world. A basic goal of digital filtering is to perform the same kind of function as does an analog filter, but it is used after sampling rather than before. In some cases, the digitally filtered signal is then converted to an analog signal. These ideas are illustrated in Figure 1.
An expert signal processor. An expert machine learner. All in one person!
I am very pleased to announce that my signal-processing, machine-learning, and modulation-recognition collaborator and friend John Snoap has successfully defended his doctoral dissertation and is now Dr. Snoap!
I started working with John after we met in the Comments section of the CSP Blog way back in 2019. John was building his own set of CSP software tools and ran into a small bump in the road and asked for some advice. Just the kind of reader I hope for–independent-minded, gets to the bottom of things, and embraces signal processing.
As we interacted over email and zoom it became clear that John was thinking of making a contribution in the area of modulation recognition, and was also interested in learning more about machine learning using neural networks. Since I had been recently engaged in hand-to-hand combat with machine learners who were, in my opinion of course, injecting more confusion than elucidation into the field, I figured this might be a friendly way for me to understand machine learning better, and maybe there would be a way or two to marry signal processing with supervised learning. So off we went.
Fast forward four years and we’ve published five papers, with a sixth in review, that I believe are trailblazing. John is that rare person that has mastered two very different technical areas: cyclostationary signal processing and deep learning. Because I believe that neural networks do not actually learn the things that we hope they will, but need not-so-gentle nudges toward learning the truly valuable things, a researcher with one foot firmly in the signal-processing world and the other firmly in the machine-learning world has a very bright future indeed.
The title of John’s dissertation is Deep-Learning-Based Classification of Digitally Modulated Signals, which he wrote as a student in the Department of Electrical and Computer Engineering at Old Dominion University under the direction of his advisor Professor Dimitrie Popescu.
Congratulations Dr. Snoap! And thank you for everything.
If a tool isn’t appropriate for your problem, don’t blame the tool. Find another one.
Let’s take a look at a recent perspectives-style paper published in the IEEE Signal Processing Magazine called “On the Concept of Frequency in Signal Processing: A Discussion [Perspectives],” (The Literature [R195]). While I criticize the paper directly, I’m hoping to use this post to provide my own perspective, and perhaps a bit of a tutorial, on the interrelated concepts of frequency, infinity, sine waves, and signal representations.
I appreciate tutorial papers in the signal-processing literature (see, for example, my positive post on Candan’s article about the Dirac delta [impulse] function), because my jaundiced view of the field is such that I think the basics, both of mathematics and communication-related signal-processing, are neglected in favor of fawning over the research flavor of the month. Over time, everybody–students, researchers, professors–is diminished because of this lack of attention to foundations.
Continue reading “Infinity, Periodicity, and Frequency: Comments on a Recent Signal-Processing Perspectives Paper ([R195])”I think of the Z transform as the Laplace transform for discrete-time signals and systems.
Previous SPTK Post: Practical Filters Next SPTK Post: Digital Filters
In this Signal Processing ToolKit post, we look at the discrete-time version of the Laplace Transform: The Z Transform.
Continue reading “SPTK: The Z Transform”For completeness, I also correct the CSPB.ML.2022 dataset, which is aimed at facilitating neural-network generalization studies.
The same random-number-generator (RNG) error that plagued CSPB.ML.2018 corrupts CSPB.ML.2022, so that some of the files in the dataset correspond to identical signal parameters. This makes the CSPB.ML.2018 dataset potentially problematic for training a neural network using supervised learning.
In a recent post, I remedied the error and provided an updated CSPB.ML.2018 dataset and called it CSPB.ML.2018R2. Both are still available on the CSP Blog.
In this post, I provide an update to CSPB.ML.2022, called CSPB.ML.2022R2.
Continue reading “CSPB.ML.2022R2: Correcting an RNG Flaw in CSPB.ML.2022”KIRK: Everything that is in error must be sterilised.
NOMAD: There are no exceptions.
KIRK: Nomad, I made an error in creating you.
NOMAD: The creation of perfection is no error.
KIRK: I did not create perfection. I created error.
I’ve had to update the original Challenge for the Machine Learners post, and the associated dataset post, a couple times due to flaws in my metadata (truth) files. Those were fairly minor, so I just updated the original posts.
But a new flaw in CSPB.ML.2018 and CSPB.ML.2022 has come to light due to the work of the estimable research engineers at Expedition Technology. The problem is not with labeling or the fundamental correctness of the modulation types, pulse functions, etc., but with the way a random-number generator was applied in my multi-threaded dataset-generation technique.
I’ll explain after the fold, and this post will provide links to an updated version of the dataset, CSPB.ML.2018R2. I’ll keep the original up for continuity and also place a link to this post there. Moreover, the descriptions of the truth files over at CSPB.ML.2018 are still valid–the truth file posted here has the same format as the truth files available on the CSPB.ML.2018 and CSPB.ML.2022 posts.
We know that ideal filters are not physically possible. Here we take our first steps toward practical–buildable–linear time-invariant systems.
Previous SPTK Post: The Laplace Transform Next SPTK Post: The Z Transform
Before we translate the Laplace transform from continuous time to discrete time, deriving the Z transform, let’s take a step back and look at practical filters in continuous time. Practical here stands in opposition to ideal as in the ideal lowpass, highpass, and bandpass filters we studied earlier in the SPTK thread.
Continue reading “SPTK: Practical Filters”We are attempting to force a neural network to learn the features that we have already shown deliver simultaneous good performance and good generalization.
ODU doctoral student John Snoap and I have a new paper on the convergence of cyclostationary signal processing, machine learning using trained neural networks, and RF modulation classification: My Papers [55] (arxiv.org link here).
Previously in My Papers [50-52, 54] we have shown that the (multitudinous!) neural networks in the literature that use I/Q data as input and perform modulation recognition (output a modulation-class label) are highly brittle. That is, they minimize the classification error, they converge, but they don’t generalize. A trained neural network generalizes well if it can maintain high classification performance even if some of the probability density functions for the data’s random variables differ from the training inputs (in the lab) relative to the application inputs (in the field). The problem is also called the dataset-shift problem or the domain-adaptation problem. Generalization is my preferred term because it is simpler and has a strong connection to the human equivalent: we can quite easily generalize our observations and conclusions from one dataset to another without massive retraining of our neural noggins. We can find the cat in the image even if it is upside-down and colored like a giraffe.
Continue reading “The Next Logical Step in CSP+ML for Modulation Recognition: Snoap’s MILCOM ’23 Paper [Preview]”The third in a series of posts on visualizing the multidimensional functions characterizing the fundamental statistics of communication signals.
Let’s continue our progression of galleries showing plots of the statistics of communication signals. So far we have provided a gallery of spectral correlation surfaces and a gallery of cyclic autocorrelation surfaces. Here we introduce a gallery of cyclic-cumulant matrices.
When we look at the spectral correlation or cyclic autocorrelation surfaces for a variety of communication signal types, we learn that the cycle-frequency patterns exhibited by modulated signals are many and varied, and we get a feeling for how those variations look (see also the Desultory CSP posts). Nevertheless, there are large equivalence classes in terms of spectral correlation. That simply means that a large number of distinct modulation types map to the exact same second-order statistics, and therefore to the exact same spectral correlation and cyclic autocorrelation surfaces. The gallery of cyclic cumulants will reveal, in an easy-to-view way, that many of these equivalence classes are removed once we consider, jointly, both second- and higher-order statistics.
Continue reading “A Gallery of Cyclic Cumulants”What do practicing engineers think of using large-language models like ChatGPT in their research, development, and writing tasks? And is there a future for humans in signal processing?
Let’s switch things up a bit here at the CSP Blog by presenting an interview on a technical topic. I interview two characters you might recall from the post on the Domain Expertise Trap: Engineers Dan Peritum and Eunice Akamai.
With the splashy entrance of large-language models like ChatGPT into everyday life and into virtually all aspects of science, engineering, and education, we all want to know how our jobs and careers could be affected by widespread use of artificial intelligence constructs like ChatGPT, Dall-E, and Midjourney. In this interview with a couple of my favorite engineers, I get a feel for how non-AI researchers and developers think about the coming changes, and of course how they view the hype, distortions, and fabrications surrounding predictions of those changes. You can find photos of the interviewees and brief biographies at the end of the post.
The interview transcript is carefully contrived lightly edited for believability clarity.
Everything is just fine.
The IEEE sent me their annual report for 2022. I was wondering how they were responding to the poor quality of many of their published papers, including faked papers and various paper retractions. Let’s take a quick look.
Another step forward in the merging of CSP and ML for modulation recognition, and another step away from the misstep of always relying on convolutional neural networks from image processing for RF-domain problem-solving.
My Old Dominion colleagues and I have published an extended version of the 2022 MILCOM paper My Papers [52] in the journal MDPI Sensors. The first author is John Snoap, who is one of those rare people that is an expert in signal processing and in machine learning. Bright future there! Dimitrie Popescu, James Latshaw, and I provided analysis, programming, writing, and research-direction support.
Continue reading “Latest Paper on CSP and Deep-Learning for Modulation Recognition: An Extended Version of My Papers [52]”The Laplace transform easily handles signals that are not Fourier transformable by introducing an exponential damping function inside the transform integral.
Previous SPTK Post: MATLAB’s resample.m Next SPTK Post: Practical Filters
In this Signal Processing ToolKit post, we look at a generalization of the Fourier transform called the Laplace Transform. This is a stepping stone on the way to the Z Transform, which is widely used in discrete-time signal processing, especially in control theory.
Continue reading “SPTK: The Laplace Transform”‘Insufficient facts always invite danger.’
Spock in Star Trek TOS Episode “Space Seed”
As most CSP Blog readers likely know, I’ve performed detailed critical analyses (one, two, three, and four) of the modulation-recognition datasets put forth publicly by DeepSig in 2016-2018. These datasets are associated with some of their published or arxiv.org papers, such as The Literature [R138], which I also reviewed here.
My conclusion is that the DeepSig datasets are as flawed as the DeepSig papers–it was the highly flawed nature of the papers that got me started down the critical-review path in the first place.
A reader recently alerted me to a change in the Datasets page at deepsig.ai that may indicate they are listening to critics. Let’s take a look and see if there is anything more to say.
Continue reading “Update on DeepSig Datasets”The cyclostationarity of frequency-shift-keyed signals depends strongly on the way the carrier phase evolves over time. Many distinct cycle-frequency patterns and spectral correlation shapes are possible.
Let’s get back to basics by looking at a large class of signals known as frequency-shift-keyed (FSK) signals. We will leave to the side, for the most part, the very large class of signals that goes by the name of continuous-phase modulation (CPM), which includes continuous-phase FSK (CPFSK), MSK, GMSK, and many more (The Literature [R188]-[R190]). Those are treated in My Papers [8], and in a future CSP Blog post.
Here we want to look at more conventional forms of FSK. These signal types don’t necessarily have a continuous phase function. They are generally easier to demodulate and are more robust to noise and interference than the more complicated CPM signal types, but generally have much lower spectral efficiency. They are like the rectangular-pulse PSK of the FSK/CPM world. But they are still used.
Continue reading “Cyclostationarity of Frequency-Shift-Keyed Signals”Final Update on “Future Posts” Poll:
So among the CSP Blog readers that voted, I think the consensus is to produce more “on brand” posts on CSP and the Signal-Processing ToolKit. Also, there is significant interest in doing CSP with GNU Radio, which I have considerable experience with, and so I’ll likely be posting some flowgraph ideas and results at some point in 2023.
Thanks everybody! (But I’ll still rant and rave from time to time; sorry!)
Update June 25, 2023: When I said you can vote multiple times, I didn’t mean to ‘spam’ the poll (as my kids would say). Someone just voted for one of the responses ten times in a row (same IP address ten votes within one minute). I meant you can vote for several different items in the poll! So I did remove some of those identical votes. I’ll close the poll at the end of the day June 30.
Update May 11, 2023: Please vote in the Reader Poll below (multiple times if you’d like) soon! As of today, CSP Applications and Signal Processing ToolKit are in the lead, with Rants and Datasets at the bottom.
The CSP Blog is rolling along here in 2023!
March 2023 broke a record for pageviews in a calendar month with over 7,000 as of this writing early in the day on March 31.
Let’s note some other milestones and introduce a poll.
What a month! We’re at about 7,145 views right now, and the previous monthly record is 6,482.
2023 was the year that a CSP Blog post crossed the 20,000-view milestone: The Spectral Correlation Function. The Cyclic Autocorrelation Function is not far behind.
About 84,000 visitors have been counted over the years since the CSP Blog launched in 2015, with 5,500 this year already. I believe this is just a count of the unique IP addresses that have accessed a page. But the number of subscribers is only 198! You can subscribe (“Follow”) to the CSP Blog by entering an email address in the “Follow Blog via Email” box on the right edge of any viewed page, near the top of the page. You’ll get notified through that email address whenever there is a new post. CSP Blog readers cannot see that email address, just as they cannot see the email address associated with any comment, unless there is an associated gravatar.
I’m planning to have more time available to devote to improving and extending the CSP Blog over the next few months. If you want to have input into that process, consider voting in the poll below.
Thanks so much to all my readers!