⚡ electrical engineering

Rethinking RSSI for WiFi Sensing

This paper introduces WiRSSI, a bistatic WiFi sensing framework that demonstrates the feasibility of using coarse Received Signal Strength Indicator (RSSI) data for accurate passive human tracking and gesture recognition by extracting Doppler, Angle-of-Arrival, and delay cues, offering a hardware-friendly alternative when Channel State Information is unavailable or restricted.

Original authors: Zhongqin Wang, J. Andrew Zhang, Kai Wu, Y. Jay Guo

Published 2026-02-25

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Zhongqin Wang, J. Andrew Zhang, Kai Wu, Y. Jay Guo

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 conversation in a noisy room.

The Old Way (CSI):
Most modern WiFi sensing research tries to listen to every single word of that conversation. They use a super-sensitive microphone (Channel State Information, or CSI) that captures the exact pitch, volume, and timing of every sound wave. It's incredibly precise, like having a high-definition recording. But there's a catch: many devices don't let you access this "high-definition" audio because of privacy laws, software locks, or hardware limitations. It's like trying to listen to a private conversation, but the door is locked.

The New Idea (WiRSSI):
This paper asks a bold question: What if we only listen to the volume?

The authors introduce WiRSSI, a system that uses only the Received Signal Strength Indicator (RSSI). In everyday terms, RSSI is just the "signal bars" on your phone. It's a very rough, low-resolution measurement. It doesn't tell you what the signal is, just how loud it is.

Traditionally, scientists thought RSSI was too "blurry" to be useful for sensing. They thought it was like trying to guess the shape of a person walking by only looking at the shadow they cast on a wall. But the authors realized that even a blurry shadow can tell you a lot if you know how to look at it.

How WiRSSI Works (The Metaphors)

Here is how they turned a "blurry shadow" into a clear tracking system:

1. The "Echo Chamber" Analogy (The Physics)
Imagine you are shouting in a canyon.

The Static Path: You hear your own voice bouncing off the far wall immediately. This is the "static" signal (the direct line between the WiFi router and the receiver).
The Moving Object: A person walks between you and the wall. Their body reflects your shout, creating a second, slightly delayed echo.
The Magic: Even though RSSI only measures the total volume of all the echoes combined, the interaction between the main shout and the person's echo creates a "beat" or a fluctuation in the volume.
The Insight: The authors realized that these volume fluctuations aren't random noise. They contain hidden clues about how fast the person is moving (Doppler), where they are standing (Angle), and how far away they are (Delay). It's like hearing the "wah-wah-wah" sound of a passing siren; even if you can't see the siren, the change in pitch tells you exactly what's happening.

2. The "Shadow Puppet" Trick (The Processing)
Since RSSI is just a single number (volume) changing over time, it's hard to tell where the person is.

The Setup: The researchers used one WiFi transmitter and a receiver with three antennas (like three ears).
The Trick: By comparing the tiny differences in volume arriving at "Ear 1," "Ear 2," and "Ear 3," they can triangulate the direction of the person.
The Math: They use a mathematical tool called a Fast Fourier Transform (FFT). Think of this as a super-fast prism. It takes the messy, fluctuating volume signal and splits it into a clear map showing "Speed" vs. "Direction." Even though the input was blurry, the output map is surprisingly sharp.

3. The "Volume-to-Distance" Guess (The Delay)
Usually, to know how far away something is, you need to know the exact time a signal took to bounce back (which requires phase information that RSSI lacks).

The Workaround: The authors realized that in a room, the further away a person is, the quieter their echo becomes (due to physics). They created a simple rule: Louder echo = Closer; Quieter echo = Further.
Calibration: They did a one-time "test run" to calibrate how loud a human echo usually is at a specific distance. After that, the system can guess the distance just by listening to the volume.

The Results: "Good Enough" is Great

The researchers tested this in a real room with people walking in circles, lines, and rectangles.

The "High-Def" System (CSI): It was the gold standard, tracking people with errors of about 0.5 meters (roughly 1.5 feet).
The "Blurry Shadow" System (WiRSSI): It tracked people with errors of about 0.8 meters (roughly 2.5 feet).

Why is this a big deal?

Privacy: RSSI doesn't capture the "fine details" of motion that could be used to identify who is moving or what they are doing in a creepy way. It's safer for privacy.
Availability: You don't need special hardware or hacked software. Every single WiFi router and phone already has RSSI. It works out of the box.
Cost: It's free. You don't need to buy new sensors; you just use the WiFi you already have.

The Bottom Line

Think of CSI as a 4K camera and RSSI as a black-and-white security camera.

The 4K camera gives you a perfect picture, but it's expensive, hard to install, and raises privacy alarms.
The black-and-white camera is grainy and low-res, but it's everywhere, cheap, and privacy-friendly.

This paper proves that with the right software, the "grainy black-and-white camera" (RSSI) can actually track people's movements and even recognize gestures (like waving or clapping) well enough for practical use. It's not about replacing the 4K camera; it's about realizing that the low-tech option is powerful enough to solve many problems when the high-tech option isn't available.

1. Problem Statement

Integrated Sensing and Communication (ISAC) aims to use wireless signals for both data transmission and environmental sensing. While Channel State Information (CSI) is the standard for fine-grained WiFi sensing, it faces significant deployment barriers:

Availability: Access to CSI is often restricted by firmware, drivers, or vendor policies on commodity devices.
Privacy: Fine-grained CSI encodes detailed motion signatures, raising privacy and compliance concerns in human-centric applications.
Hardware Limitations: CSI-based sensing is sensitive to hardware-induced phase distortions (e.g., clock asynchrony, PLL jumps).

Conversely, the Received Signal Strength Indicator (RSSI) is ubiquitously available on all WiFi devices but is widely regarded as too coarse for fine-grained sensing. It is typically a bandwidth-aggregated, low-resolution (8–10 bits) power measurement that discards multipath diversity and subcarrier-level phase information, leading to the belief that it cannot capture micro-motion or precise geometry.

The Core Question: Can RSSI, traditionally viewed as inadequate, be repurposed for passive human tracking and motion sensing if processed with the right physical-layer models?

2. Methodology: The WiRSSI Framework

The authors propose WiRSSI, a bistatic WiFi sensing framework that enables passive human tracking using RSSI-only measurements. The system employs a 1Tx-3Rx configuration (one transmitter, three receiving antennas) and relies on a lightweight signal processing pipeline.

A. Theoretical Foundation: RSSI as Aggregated CSI Power

The paper derives a physical model showing that RSSI is not merely a path-loss indicator but implicitly preserves motion information.

Signal Model: RSSI is modeled as the aggregated power of CSI across all subcarriers.
Cross-Term Preservation: When CSI power is calculated ( $|CSI|^2$ ), random phase distortions (Clock Offset, CFO, PO) are eliminated. However, the cross-term between the static path (Line-of-Sight) and the dynamic path (target reflection) remains.
Implicit Encoding: This cross-term encodes the interaction between the static and dynamic paths, preserving Doppler shifts, Angle-of-Arrival (AoA), and delay information within the power domain, despite the lack of explicit phase data.

B. Processing Pipeline

Static Clutter Removal: RSSI is converted to a linear scale. The temporal mean is subtracted to suppress the dominant static component (LOS/NLOS from walls), isolating the dynamic motion-induced variations.
Doppler-AoA Estimation (2D FFT):
- A Temporal FFT is applied to extract Doppler frequency shifts.
- A Spatial FFT is applied across the three antennas to estimate the Angle-of-Arrival (AoA).
- Ambiguity Resolution: Since RSSI is real-valued, the Doppler spectrum is symmetric. The system resolves this mirror ambiguity by leveraging the known geometric direction of the transmitter ( $\theta_S$ ), effectively breaking the symmetry in the joint Doppler-AoA spectrum.
Delay Estimation (Amplitude-Based):
- Unlike CSI, RSSI lacks subcarrier phase for direct delay estimation.
- WiRSSI exploits the amplitude-delay relationship: The amplitude of the reflected signal is inversely proportional to the propagation distance (path loss).
- By calculating the ratio between the dynamic path amplitude and the static path amplitude, and using a pre-calibrated reflection coefficient ratio ( $\gamma$ ), the system infers the bistatic delay (distance).
Tracking & Smoothing: Estimated polar coordinates (AoA, Delay) are converted to Cartesian coordinates. A Hampel filter removes outliers, followed by a Savitzky-Golay (SG) filter to smooth the trajectory, ensuring robustness against RSSI noise.

3. Key Contributions

Physical-Layer Reinterpretation: The paper proves that RSSI implicitly encodes phase-related motion information (Doppler, AoA, Delay) through the interaction of static and dynamic paths in the power domain, challenging the notion that RSSI is useless for fine-grained sensing.
WiRSSI Framework: A novel, low-complexity framework that extracts Doppler-AoA features via 2D FFT and estimates delay via amplitude modeling, enabling passive tracking without subcarrier-level phase data.
Hardware-Friendly Implementation: The system runs on commodity Intel 5300 NICs using only RSSI streams, requiring no firmware modifications or driver-level access to CSI.
Comprehensive Evaluation: Extensive experiments comparing WiRSSI against state-of-the-art CSI-based methods (SRCC, CASR) and mmWave radar ground truth.

4. Experimental Results

Experiments were conducted in practical indoor environments with human subjects moving along elliptical, linear, and rectangular trajectories.

Tracking Accuracy:
- WiRSSI (RSSI-only): Achieved median XY localization errors of 0.905 m (elliptical), 0.784 m (linear), and 0.785 m (rectangular).
- CSI Baselines: A representative CSI-based method achieved errors of 0.574 m, 0.599 m, and 0.514 m respectively.
- Conclusion: While CSI is more accurate, RSSI achieves sub-meter median accuracy, which is sufficient for many practical ISAC applications.
Gesture Recognition:
- Evaluated on the Widar3.0 dataset.
- WiRSSI features achieved 75.3% accuracy in gesture recognition (vs. 93.9% for CSI-based SRCC).
- Crucially, in a strict cross-domain generalization test (training on one location, testing on others), WiRSSI maintained 50.7% accuracy, outperforming several CSI-based baselines and demonstrating robustness to domain shifts.
Computational Efficiency:
- WiRSSI processes a Coherent Processing Interval (CPI) in 0.2 ms, compared to 12 ms for the CSI-based SRCC method. This makes WiRSSI significantly more suitable for real-time, resource-constrained edge devices.
Robustness: The system remains functional even with downsampled RSSI rates (tens of Hz) and varying Tx-Rx distances (2–4m optimal).

5. Significance and Implications

Democratizing WiFi Sensing: WiRSSI provides a viable sensing solution for environments where CSI is restricted, unreliable, or privacy-sensitive. It turns a "coarse" metric into a functional sensing tool.
Privacy-Preserving: Since RSSI does not expose fine-grained phase signatures, it is less invasive for privacy-sensitive applications (e.g., healthcare monitoring in private spaces).
Cost-Effective Deployment: It leverages existing commodity hardware without requiring specialized sensors or firmware hacks, facilitating large-scale ISAC deployment.
Hybrid Potential: The work suggests a future where RSSI and CSI can be used in hybrid schemes, with RSSI serving as a fallback or complementary modality to enhance system robustness.

In summary, the paper successfully redefines RSSI from a simple signal strength indicator to a capable sensing modality, offering a practical, low-cost, and privacy-friendly alternative to CSI for human tracking and motion sensing.