Automated model for analysis of .IQ and .wav files along with signal parameter extraction
National Technical Research Organisation (NTRO) · Miscellaneous · Software
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.
What it actually is
Radio signals captured off the air are stored as raw IQ or WAV recordings, and analysts currently work out their parameters by hand. The ask is an automated tool that reads these files and extracts signal characteristics — modulation type, sampling rate, forward error correction, interleaving and so on — and can demodulate the signal, through a GUI.
What to build
A GUI signal-analysis tool that loads IQ and WAV recordings, handling the different internal formats of each, computes and displays the spectrum and spectrogram, and automatically extracts signal parameters — estimating symbol rate, identifying the modulation type through a classifier or feature-based analysis, and detecting characteristics like FEC and interleaving where feasible — then demodulates the signal to recover the underlying data, built on GNU Radio and Python signal-processing tooling with a display for waveform, spectrum and extracted parameters.
Smallest thing that wins the room
Load an IQ recording of a known modulation, show the tool computing the spectrum, correctly classifying the modulation type and estimating the symbol rate, then demodulating it to recover the bitstream.
How crowded this one gets
A guess, projected from the 2025 statements — the last year where both the submission counts and the winners were published.
Quieter than 39% of the 226 · #139 of 226 by expected field
A normal-sized field. Your idea has to be good, not miraculous.
Why: defence, intelligence and space bodies drew small fields.
This is a guess, not a fact
Nobody has published 2026’s numbers yet. This is an analysed estimate from last year’s pattern, so please do not take it as the truth — check the live counter on the SIH portal before you decide anything. The range covers the middle half of likely outcomes, so one statement in two lands outside it. Entry closes at 500 ideas per statement, so no range goes past that — a statement that reaches the cap fills and shuts rather than drawing an unlimited crowd. The model reads only three things a team can see before choosing — software or hardware, the theme, and what kind of body posted it — and those explain about a quarter of the variation in last year’s field sizes (R² 0.25 on held-out statements). Trust the band more than the number, and the ordering more than either. It cannot see how good your idea is, which is the part that actually decides it.
The scores
The number is the shorthand. The line under it is the reason.
Acceptance potential
2/5Modulation 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.
Feasibility
2/5Loading 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
3/5Automatic 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
3/5The 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.
Effort
MassiveFile 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
MediumClassifying 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.
In its favour
- Green flag: Public RF modulation datasets like RadioML give you labelled signals to train and validate a modulation classifier on
- Green flag: GNU Radio provides mature building blocks for spectral analysis and demodulation, so you assemble rather than implement from scratch
- Green flag: A clean modulation-classification-and-demodulation demo on a known signal is genuinely impressive
- Green flag: The specialist RF domain guarantees a thin field
Against it
- Red flag: 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
- Red flag: Blind demodulation of an arbitrary unknown signal requires parameter estimation that even professional tools do semi-manually, so the automated claim is overreach
- Red flag: The description is loosely scoped, so a team can build modulation classification and still have answered only a fraction of what was asked
- Red flag: NTRO evaluators do this professionally and will test the tool on a signal it has not seen, where it will likely fail
What you will be writing
- GNU Radio flowgraphs + Python
- IQ / WAV file parsing
- Automatic modulation classification (CNN on spectrogram / feature-based)
- Symbol-rate estimation (cyclostationary analysis)
- Demodulation chains per modulation type
- PyQt GUI with spectrum/spectrogram display
- Signals intelligence
- RF signal processing
- Automatic modulation classification
Prior art to read before you start
automatic modulation classification · signal parameter extraction from IQ · blind demodulation
Analysed by Claude Opus. Every score above is a judgment call with its reasoning attached — kindly cross-check this against the official statement on the SIH portal before your team commits to it.