Conventional approaches to unknown waveform detection have generally fallen into two camps. Statistical anomaly detection methods build a model of what normal signals look like and flag anything that deviates from it, an approach that is conceptually elegant but notoriously brittle in cluttered spectral environments where legitimate signals vary wildly in power, modulation and bandwidth. Deep learning classifiers, meanwhile, have delivered impressive results on benchmark datasets of known modulations, yet they stumble when confronted with waveform families absent from their training sets. Some researchers have tried to bridge this gap with generative techniques that synthesize artificial examples of unknown signals, but generating realistic synthetic samples is itself an unsolved problem, and poorly generated data can bias a model in ways that are difficult to diagnose. The result, in practice, is that fielded systems often fail exactly when they are needed most: during encounters with genuinely novel emitters.

The new discriminative model sidesteps synthetic sample generation altogether. Instead of trying to imagine what unknown waveforms might look like, it learns a richer description of the signals it does know, so that anything sufficiently different stands out sharply. The key architectural decision is the joint use of time-domain and frequency-domain representations. The time domain captures how a signal’s amplitude evolves moment to moment, preserving transient features, timing structure and modulation transitions that unfold in sequence. The frequency domain, obtained through transforms such as the Fourier transform, reveals how energy is distributed across the spectrum, exposing carrier offsets, spectral occupancy, harmonic structure and bandwidth fingerprints. Human signal analysts have long toggled between these two views on oscilloscopes and spectrum analyzers; the invention encodes that dual perspective directly into the learning pipeline.

But merely concatenating two views of a signal would not, by itself, guarantee better detection. The second pillar of the invention is a cosine similarity loss function that reshapes the model’s internal feature space. In machine learning, a loss function defines what a model is penalized for getting wrong, and therefore what it learns to prioritize. Cosine similarity measures the angle between two feature vectors rather than the distance between them, meaning it is sensitive to the direction of a representation but insensitive to its magnitude. By training with a cosine similarity objective, the system is pushed to align the feature vectors of signals from the same class more tightly while steering vectors of different classes apart. Class-specific features become more cleanly separated, and the model develops a sharper decision boundary between familiar waveform families and everything else.

The practical consequence of this design is a measurable jump in performance. In testing against comparable models that lacked the combined time-frequency representation and the cosine similarity mechanism, the invention delivered approximately a 10 percent improvement in detection accuracy for unknown waveforms. A ten percent gain may sound incremental, but in the context of unknown-signal detection, where baseline systems frequently operate in regimes of unreliable performance, it represents a substantial margin. It means fewer missed detections of genuinely anomalous emitters and fewer false alarms triggered by ordinary variations in known signals, both of which carry real operational costs. The improvement stems directly from the model’s enhanced ability to differentiate subtle waveform variations that do not appear in its training data, rather than from any increase in raw computational capacity.

Robustness and generalization are the qualities that make this improvement durable rather than dataset-specific. Because the model’s features are aligned by direction in the feature space, they are less sensitive to the scale of a signal’s power, a property that matters enormously in realistic radio environments where the same emitter may be received at wildly different strengths depending on distance, terrain and antenna orientation. The dual-domain representation also provides redundancy: a waveform feature that is ambiguous in the time domain, such as a slight shift in spectral occupancy, may be unmistakable in the frequency domain, and vice versa for temporal phenomena. This built-in cross-checking gives the system a form of resilience that single-representation classifiers lack, allowing it to maintain accurate classification under the noisy, adversarial conditions that characterize contested spectrum.

The anticipated applications span both military and civilian domains. In electronic warfare, the ability to detect and classify unknown communication signals is foundational to situational awareness, since an adversary’s new emitter is by definition absent from any pre-existing threat library. Spectrum management authorities could deploy the technology to monitor and enforce the use of the electromagnetic spectrum, identifying rogue transmissions and interference sources that conventional monitoring tools miss. Intelligence, surveillance and reconnaissance operations depend on reliable waveform identification, and the invention’s resistance to the biases introduced by synthetic sample generation makes it a more trustworthy analytic tool. Radio astronomers, who fight a constant battle against radio-frequency interference contaminating observations of faint cosmic sources, could use the model to detect and mitigate intruding signals. Communication security systems, meanwhile, could apply it to flag unauthorized or anomalous transmitters operating within protected networks.

The technology is at technology readiness level 3, the stage at which a concept has been proven analytically and experimentally in laboratory conditions but has not yet been integrated into an operational prototype. It is patent pending and available for licensing through the Research Foundation for the State University of New York, the nation’s largest research foundation, which supports research across the SUNY system in areas including artificial intelligence for the public good, quantum technologies, next-generation semiconductors, biotech and medicine, and energy and climate solutions. SUNY, the largest comprehensive system of higher education in the United States, oversees nearly a quarter of academic research in New York, with research expenditures of nearly 1.5 billion dollars in fiscal year 2025, and the foundation offers multiple pathways for translating such innovations into commercial and economic development opportunities.

What makes the invention notable beyond its immediate applications is the clarity of its underlying insight. Much of modern machine learning progress has come from adding scale: more layers, more parameters, more data. This work demonstrates that careful attention to how a signal is represented, and to how a model is trained to organize its internal features, can yield decisive gains without any of that overhead. By combining two complementary mathematical descriptions of the same physical phenomenon and aligning them with a geometry-aware loss function, the researchers have built a system that sees the electromagnetic world the way an experienced analyst does, from multiple angles at once. As the radio spectrum grows more crowded and the population of uncooperative or unrecognized transmitters continues to expand, tools that can reliably distinguish the known from the genuinely unknown will only grow in importance, and this time-frequency approach offers a concrete, testable step toward that goal.

Subject of Research: An AI model combining time-domain and frequency-domain features with cosine similarity loss for detecting unknown electromagnetic waveforms.

Article Title: Contrasting time-frequency representations for unknown waveform detection

Article References: Contrasting time-frequency representations for unknown waveform detection. (n.d.). Original publication

Image Credits: AI Generated

DOI: Not provided

Keywords: unknown waveform detection, cosine similarity loss, time-frequency analysis, electromagnetic spectrum, machine learning, electronic warfare, spectrum management, radio-frequency interference, signal classification, deep learning, radio astronomy, SUNY licensing

Each of the four ingredients addresses a specific bottleneck of the ultra-massive machine-type communications scenario that is expected to define 6G. Intelligent reflecting surfaces, or IRS, are essentially walls of tiny, nearly passive elements that can be tuned to steer radio waves in chosen directions, boosting signal strength along useful paths and shielding receivers from interference. Because they involve no power-hungry amplifiers or radio chains, they offer a cheap way to reshape the propagation environment around energy-constrained Internet of Things devices. Simultaneous wireless information and power transfer, or SWIPT, exploits the same radio signals to deliver both data and energy, allowing a receiving device to split the incoming wave so that part of it is decoded as information and part is harvested as electrical power. For sensors that cannot easily be recharged or wired to the grid, this dual function can be transformative.

The other two technologies attack the problem from the network side. NOMA allows several users to share the same time, frequency and code resources by superimposing their signals at different power levels and separating them at the receiver through successive interference cancellation, dramatically improving spectral efficiency. Full-duplex transmission lets a node transmit and receive simultaneously on the same frequency, in principle doubling spectral efficiency, although it must contend with powerful self-interference from its own transmitter. Combining these elements, the proposed scheme creates a full-duplex NOMA system in which a base station serves a downlink user while, at the same time, energy-constrained uplink devices harvest power from the downlink signal, assisted by an IRS that strengthens both directions of the exchange. Earlier work by the same authors had demonstrated the promise of such a combination for 6G ultra-massive machine-type communications under ideal assumptions; the new study confronts the harder, more realistic case.

That harder case is imperfect channel state information, or CSI. Every intelligent beamforming and resource allocation decision in an IRS-assisted system depends on knowing precisely how signals travel from transmitter to surface and from surface to receiver. Yet IRS panels contain no active components, which means there is no easy way to insert pilots or measure the cascaded channels element by element. Estimating the channel of each individual reflecting element would impose a crushing signalling overhead, so practical systems must work with estimates that carry error. If a design assumes perfect CSI and reality deviates even slightly, the promised quality of service for the downlink user can quietly collapse, and the harvested energy that uplink devices count on may never arrive. Robust design, in which performance guarantees hold for the worst possible channel error within a bounded uncertainty set, is therefore not a luxury but a necessity for IRS-assisted networks.

The researchers formulated the core problem as a worst-case uplink sum rate maximization: the goal is to maximize the total data rate achieved by the uplink NOMA devices in the most pessimistic allowable channel condition, subject to the constraint that the downlink user’s quality of service is always satisfied. This is a notoriously difficult non-convex optimization problem, made harder by the coupling between the beamforming vectors at the base station, the phase shifts applied by the IRS elements, and the power-splitting ratios that divide each received signal between energy harvesting and information decoding. Solving it exactly is mathematically intractable, so the authors constructed an alternating optimization algorithm that breaks the problem into manageable pieces, optimizing one set of variables at a time while holding the others fixed and iterating until convergence.

Each iteration relies on an arsenal of modern optimization techniques. The S-Procedure converts infinitely many worst-case constraints, one for every possible channel error, into tractable matrix inequalities. The penalty convex-concave procedure and successive convex approximation linearize the non-convex pieces of the problem around the current solution, yielding a sequence of convex subproblems that can be solved efficiently with standard tools. The Schur complement rewrites determinant-based matrix inequalities into forms that standard solvers can handle, while Generalized Petersen’s Lemma helps convert products of uncertain channel terms into expressions where the uncertainty can be bounded cleanly. Together, these methods produce a solution that is not merely a good guess, but one with built-in protection: no matter how the true channels wander within the assumed error bounds, the quality of service and the harvested energy requirements are guaranteed to hold.

The numerical studies accompanying the paper show that this robustness pays measurable dividends. Compared with benchmark schemes, the proposed design achieves a higher worst-case uplink sum rate, demonstrating that the joint tuning of IRS phase shifts, transmit beamforming and power splitting extracts more from the same spectrum and the same radiated power. More importantly, the robust scheme provides more practical and dependable results than its non-robust counterpart. When channel errors are injected into the simulations, designs built on perfect-CSI assumptions suffer degraded performance and occasional violations of the downlink quality-of-service constraint, whereas the robust design keeps its guarantees intact. The gap between the two widens as the uncertainty grows, underscoring that the extra mathematical machinery is not academic hair-splitting but the difference between a system that works in the lab and one that works in the field.

The implications reach well beyond one particular algorithm. The ultra-massive machine-type communications scenario envisions connection densities and energy constraints far beyond what cellular networks have ever handled, and the study argues that no single technology can meet them alone. IRS supplies coverage and signal gain without adding active infrastructure, SWIPT turns the network’s own transmissions into a power grid for its sensors, NOMA squeezes many connections into scarce spectrum, and full-duplex operation doubles the utility of every radio resource. Stacking the four, and making the stack resilient to imperfect channel knowledge, sketches a plausible blueprint for the machine-centric fabric of 6G, in which thousands of self-powered devices communicate reliably through intelligently steered reflections.

There remain formidable engineering hurdles before such systems reach commercial deployment, from scalable channel estimation for large IRS panels to the hardware cost of full-duplex self-interference cancellation. But by formulating the problem with bounded channel errors from the outset and proving that a computationally feasible algorithm can still maximize worst-case throughput, the researchers have moved the field closer to designs that survive contact with reality. For a 6G vision built on trillions of devices, that kind of robustness may prove to be the most valuable specification of all.

Subject of Research: A robust IRS- and SWIPT-assisted full-duplex NOMA transmission scheme designed for 6G ultra-massive machine-type communications with imperfect channel state information.

Article Title: Robust IRS and SWIPT Assisted Full-Duplex NOMA for 6G Networks

Article References: Goktas, M. B., Al-Basit, S. M., & Ding, Z. (2026). Robust IRS and SWIPT Assisted Full-Duplex NOMA for 6G Networks. Journal of Network and Systems Management, 34(4), Article 131. https://doi.org/10.1007/s10922-026-10099-6

Image Credits: AI Generated

DOI: 10.1007/s10922-026-10099-6

Keywords: 6G, intelligent reflecting surface, SWIPT, full-duplex communications, NOMA, imperfect channel state information, Internet of Things, ultra-massive machine-type communications, beamforming optimization, wireless power transfer, convex optimization, sum rate maximization