Automated model for analysis of .IQ and .wav files along with signal parameter extraction
National Technical Research Organisation (NTRO)
Bold. Let us find out precisely how bold, in the order a panel will find out.
Proceed with caution. Scope this to automatic modulation classification and symbol-rate estimation, which are achievable, and be explicit that FEC, interleaving and blind demodulation are beyond a hackathon — promising the full wish list to an NTRO judge is a claim the tool cannot back up. Roughly 150–340 teams are expected to go here.
The receipts
Every red flag on this statement, in full. These are the four places it bites.
Exhibit A
Detecting FEC scheme and interleaving from a raw recording is an unsolved-in-general SIGINT problem, so promising it invites a demonstration you cannot deliver
It gets worse
Blind demodulation of an arbitrary unknown signal requires parameter estimation that even professional tools do semi-manually, so the automated claim is overreach
Still reading?
The description is loosely scoped, so a team can build modulation classification and still have answered only a fraction of what was asked
And the finisher
NTRO evaluators do this professionally and will test the tool on a signal it has not seen, where it will likely fail
The damage report
Every score this statement earned, and what each one actually costs you.
Feasibility
2/5You have picked a fight with physics, procurement, or both. One of them always wins.
Loading files, spectra and basic modulation classification are achievable, but automatically extracting FEC scheme and interleaving from a raw recording is a genuinely hard signals-intelligence problem, and demodulating arbitrary unknown signals requires estimating parameters that professional SIGINT tools struggle with — this is deep RF domain expertise, not a weekend build.
Innovation scope
3/5Mildly interesting. The novelty will not carry the room; the build has to.
Automatic modulation classification is an established research area and the description names the tools, so the approach is somewhat constrained, though the parameter-extraction depth leaves room.
Clarity
3/5Clear enough to start, vague enough to drift. Write the scope down and stop reinterpreting it weekly.
The parameters to extract are named but the description is loosely worded, does not specify which signal types must be handled or to what accuracy, and conflates a broad wish list with the achievable core.
Acceptance potential
2/5The numbers do not like you. Bring something the numbers cannot see.
Modulation classification on clean known signals is doable, but the FEC, interleaving and general demodulation the description asks for is professional SIGINT territory, and NTRO judges who do this work will immediately see the gap between a demo on a known signal and the real analysis of an unknown one.
Effort
MassiveA semester of work wearing a hackathon costume. Something is getting cut; decide what now, not in week five.
File handling, spectral analysis, modulation classification, symbol-rate estimation, FEC and interleaving detection, and demodulation across signal types is a large signals-intelligence toolchain.
Demo-ability
MediumDemoable, if you rehearse it. Nobody rehearses it.
Classifying a modulation and demodulating to recover bits is a satisfying result on a known signal, but on an arbitrary unknown recording the tool will often fail, and the interesting cases are exactly the hard ones.
Data
None suppliedNo dataset comes with this one, so every accuracy figure you quote is a number about labels you invented.
Nothing is provided with the statement. You are sourcing, cleaning and labelling it yourself, and that work is invisible in the demo but very visible in the questions.
The demo they will have already seen
Somewhere around 150–340 teams are heading here, and the description is doing the choosing for most of them. They will read the same brief, reach the same architecture, and build a version of the same demo you are planning. Being correct is the floor. If your five minutes could be swapped with the team before you and nobody in the room would notice, you have not picked badly — you have built predictably, which costs exactly the same and hurts more.
What survives
The ground worth standing on when the questions start.
- Public RF modulation datasets like RadioML give you labelled signals to train and validate a modulation classifier on
- GNU Radio provides mature building blocks for spectral analysis and demodulation, so you assemble rather than implement from scratch
- A clean modulation-classification-and-demodulation demo on a known signal is genuinely impressive
None of that means do not pick it. It means do not walk into that room having heard any of this for the first time from a judge.
The framing is a joke. The findings are not — they are the same analysis on the statement page, and every line above is attached to a score or a fact in the record. It is one opinion with its reasoning attached, so argue with it before you trust it.