⚡ electrical engineering

Elliptic Range-Doppler Mapping for OFDM-ISAC under IQ Imbalance

This paper proposes an elliptic group orthogonal matching pursuit detector that directly exploits the coupled structure of IQI-impaired OFDM-ISAC observations to improve range-Doppler recovery and weak-target detection without prior IQI compensation.

Original authors: Thrassos K. Oikonomou, Dimitrios Bozanis, Sotiris A. Tegos, Panagiotis D. Diamantoulakis, George K. Karagiannidis

Published 2026-07-17

📖 7 min read🧠 Deep dive

CC BY 4.0

Original authors: Thrassos K. Oikonomou, Dimitrios Bozanis, Sotiris A. Tegos, Panagiotis D. Diamantoulakis, George K. Karagiannidis

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are trying to listen to a friend whispering a secret across a noisy room. In the world of wireless technology, this "room" is the airwaves, and the "whisper" is a signal carrying both data and a map of the surroundings. This is the heart of a field called Integrated Sensing and Communication (ISAC). Instead of using one device to talk and another to look around (like a phone and a radar), ISAC tries to do both with the same signal, like a single flashlight that illuminates a room while also sending a message.

To make this work, engineers use a clever trick called OFDM (Orthogonal Frequency Division Multiplexing). Think of OFDM as a giant piano keyboard where each key is a different frequency. By playing many keys at once, the system can send lots of data and, by listening to how the sound bounces back, figure out where objects are and how fast they are moving. This "map" of distance and speed is called a Range-Doppler map. However, just like a cheap microphone might distort a singer's voice, the hardware in our devices often has a flaw called "IQ Imbalance." This glitch mixes up the signal's "left" and "right" channels, creating confusing "ghost" echoes that look like fake targets on the map. If you try to fix this by cleaning the audio first and then looking for the targets, you often make the noise worse or miss the quiet whispers entirely.

This paper tackles that exact problem. The authors, a team from the Aristotle University of Thessaloniki, propose a new way to listen to these distorted signals. Instead of trying to "fix" the hardware glitch before looking for targets, they decided to embrace the mess. They realized that a real target doesn't just create one echo; it creates a pair of echoes—a direct one and a "mirror" one—that are mathematically linked. They call this pair an "elliptic atom." By treating these two linked echoes as a single, unified shape (an ellipse) rather than two separate problems, they built a new detector called "Elliptic GOMP."

In their simulations, this new method proved to be much better at finding the real targets and ignoring the ghosts, especially when the hardware distortion was strong or when there were many targets crowded together. It even managed to spot a very weak target hiding right next to a loud, strong one, a task where older methods failed. The researchers showed that by understanding the specific shape of the distortion, they could turn a hardware flaw into a useful clue, finding the truth in the noise without needing to perfectly clean the signal first.

The Story of the Ghost and the Mirror

Imagine you are playing a game of "Marco Polo" in a dark gymnasium. You shout "Marco!" and wait for the "Polo" to bounce back. In a perfect world, you hear one clear voice coming from a specific direction, telling you exactly where your friend is. But imagine the gym has a weird, warped wall that splits every sound you make into two: a clear voice and a ghostly echo that sounds slightly different and comes from a slightly different angle.

In the world of wireless signals, this "warped wall" is the IQ Imbalance. It happens because the electronic circuits inside our receivers aren't perfectly balanced. When a signal bounces off a car or a person, the receiver doesn't just see the real object; it also sees a "ghost" version of it. This ghost is a mirror image: if the real car is moving forward, the ghost seems to be moving backward. If the real car is 10 meters away, the ghost is also 10 meters away.

For a long time, engineers tried to solve this by building a "noise-canceling headphone" for the signal. They would try to mathematically remove the ghost before they tried to find the real car. The paper argues that this is like trying to fix a blurry photo by sharpening the pixels first, only to realize that the blur actually contains a clue about the object's shape. The old methods (like OFDM-ZF or LMMSE) try to force the signal back into a "perfect" shape, but in doing so, they often amplify the noise or miss the faintest details.

The "Elliptic" Insight

The authors of this paper had a "Eureka!" moment. They realized that the real target and its ghost aren't two separate enemies; they are two sides of the same coin. They are locked together by the laws of physics. If you know the strength of the real target, you automatically know the strength of the ghost, just flipped and mirrored.

They decided to stop fighting the ghost and start dancing with it. They created a new mental model called an "Elliptic Atom."

Think of a normal target as a perfect circle. If you spin a circle, it looks the same from every angle. But when the "IQ Imbalance" hits, that circle gets squished into an ellipse (like a stretched-out oval). The real target is one side of the oval, and the ghost is the other. The paper suggests that instead of trying to squash the oval back into a circle (which is hard and noisy), we should just look for the oval itself.

They built a detective tool called Elliptic GOMP (Group Orthogonal Matching Pursuit). Imagine a search party looking for a lost hiker.

The Old Way: The search party tries to clean up the static on their radios first. If the static is too loud, they can't hear the hiker. If they try to guess the hiker's location based on a single, distorted voice, they might get lost.
The New Way: The search party knows that the lost hiker's voice will always be accompanied by a specific echo. They don't try to remove the echo; they listen for the pattern of the voice and the echo together. They look for the "oval" shape of the sound.

What They Found

The authors ran thousands of computer simulations to test their idea. They set up a virtual world with a grid of 1,024 possible spots where a target could hide (imagine a chessboard with 1,024 squares). They introduced different levels of "hardware sickness" (IQ Imbalance) and different numbers of targets.

Here is what their simulations showed:

Better at Finding the Truth: When the hardware was very "sick" (strong IQ Imbalance), the old methods got confused and started seeing ghosts. The new Elliptic GOMP method, however, kept its cool. It found the real targets much more often, even when the noise was high.
Crowded Rooms: When they put 10 targets in the grid, the old methods started to fail as the targets got closer together. The new method kept finding all 10, proving it can handle busy scenes without getting mixed up.
The "Whispering" Target: This was the most impressive test. They placed a very loud, strong target next to a very weak, quiet one. The loud target's "ghost" usually hides the quiet one. The old methods couldn't hear the quiet target at all. The Elliptic GOMP, by correctly subtracting the loud target's entire oval shape (both the real part and the ghost part), revealed the quiet target hiding right next to it.

The paper doesn't claim to have built a physical device yet; these results are from computer simulations. However, the math is solid. They showed that by using two special mathematical operations (called weighted 2D FFTs) to calculate the "oval" shapes, their method is fast enough to be used in real-time.

Why This Matters

This research suggests a shift in how we think about technology flaws. Instead of seeing a hardware imperfection as a problem to be erased, we can see it as a pattern to be understood. By treating the "ghost" and the "real" signal as a single, linked team, we can build smarter sensors that work better in the real world, where perfect hardware doesn't exist. It's a reminder that sometimes, to find the truth, you don't need to clean up the noise—you just need to learn how to listen to the whole song.

Technical Summary: Elliptic Range-Doppler Mapping for OFDM-ISAC under IQ Imbalance

Problem Statement
In Orthogonal Frequency Division Multiplexing Integrated Sensing and Communication (OFDM-ISAC) systems, receiver in-phase/quadrature imbalance (IQI) introduces significant impairments. IQI couples each OFDM subcarrier with its mirror counterpart, generating "ghost" targets and degrading range-Doppler recovery. Conventional approaches typically attempt to compensate for IQI before performing standard range-Doppler processing (e.g., via Zero-Forcing or LMMSE estimation). However, these methods often fail to exploit the inherent structure of the IQI-impaired observation. Specifically, a single physical target does not merely create an arbitrary distortion; it generates two coupled components—a direct branch and a mirror-conjugate branch—linked by the target's complex reflection coefficient and its conjugate. Existing sparse recovery methods, which rely on single-atom delay-Doppler dictionaries, do not account for this specific coupling, leading to performance degradation, particularly under moderate to strong IQI conditions.

Methodology
The paper proposes a novel framework that bypasses explicit IQI compensation in favor of direct sparse recovery on the IQI-impaired observation grid. The core methodology involves three main steps:

Elliptic Atom Group Representation: The authors demonstrate that the IQI-impaired response of a physical target at delay-Doppler cell $\theta = (\tau, \nu)$ is not a single Fourier atom but a structured group. This group consists of a direct branch and a mirror branch (with reversed Doppler phase). Because these branches are controlled by the same reflection coefficient $\alpha$ and its conjugate $\alpha^*$ , the target response traces an elliptic trajectory in the real observation space rather than a circular one. Consequently, the target is modeled as a real two-dimensional subspace (an "elliptic atom group") spanned by two real basis vectors derived from the direct and mirror branches.
Elliptic Group Orthogonal Matching Pursuit (GOMP): A detector is formulated to perform sparse recovery directly on the received grid using these elliptic groups. Unlike conventional Orthogonal Matching Pursuit (OMP) which matches scalar Fourier atoms, the proposed Elliptic GOMP performs group projections onto the two-dimensional elliptic subspace of each candidate cell. At each iteration, the algorithm selects the cell whose elliptic group captures the maximum residual energy, effectively detecting the complete physical target (direct + mirror) rather than isolated components.
Efficient Computation: To avoid the high complexity of explicitly constructing the full elliptic dictionary, the required correlations are computed efficiently. The method utilizes two weighted two-dimensional Fast Fourier Transforms (FFTs) to calculate the correlations for the direct and mirror branches. These are followed by local group projections. The computational complexity scales as $O(KMN \log(MN))$ , preserving the leading FFT order of conventional sparse OFDM processing.

Key Contributions

Structural Modeling: The paper establishes that receiver IQI transforms a physical target into a coupled "direct-plus-mirror" elliptic atom group, rather than an isolated delay-Doppler atom.
Direct Sparse Recovery: It introduces an Elliptic GOMP detector that operates directly on IQI-impaired observations, eliminating the need for a preliminary IQI compensation stage that might amplify noise or distort weak signals.
Algorithmic Efficiency: The proposed detector computes necessary correlations via weighted 2D-FFTs and local projections, maintaining computational efficiency comparable to standard sparse recovery methods.
Performance Gains: The method is shown to improve exact support recovery and weak-target detection, particularly in scenarios with moderate to strong IQI, high target density, and near-target weak targets.

Numerical Results
Simulations were conducted using an OFDM-ISAC frame with 16 symbols and 64 subcarriers, comparing the proposed Elliptic GOMP against benchmarks: OFDM-ZF + OMP, Pairwise IQI-ZF + OMP, and LMMSE + OMP.

SNR Performance: Under both moderate ( $|\lambda/\mu|=0.5$ ) and strong ( $|\lambda/\mu|=0.8$ ) IQI, the proposed method achieved reliable exact recovery at lower Signal-to-Noise Ratios (SNR) than all benchmarks. The performance gap widened significantly under strong IQI, where compensation-based methods suffered from noise enhancement and residual artifacts.
Target Density: As the number of targets increased, compensation-first methods degraded rapidly due to accumulated IQI artifacts and incorrect atom selections. The Elliptic GOMP maintained high exact recovery probabilities across a wide range of target densities.
Weak Target Detection: In scenarios where a weak target was placed adjacent to a strong target, the proposed method successfully exposed the weak target by removing the strong target's coupled direct-and-mirror response. Benchmarks failed to detect the weak target reliably, especially under strong IQI, as the residual response of the strong target masked the weak one.

Significance
The paper claims that its primary significance lies in shifting the paradigm from treating the IQI-induced mirror branch solely as an impairment to be suppressed, to utilizing it as a useful sensing structure. By jointly modeling the direct and mirror components as a single elliptic signature, the proposed method achieves superior robustness and detection accuracy without the noise amplification associated with explicit compensation. The results suggest that direct sparse recovery using receiver-IQI-aware elliptic atoms is a more effective strategy for OFDM-ISAC in the presence of hardware impairments than traditional compensation-then-detection approaches.

The Story of the Ghost and the Mirror

The "Elliptic" Insight

What They Found

Why This Matters

More like this