LiveSense: Turning Your Laptop's Wi-Fi Into a Centimeter-Precision Radar

Paper: 2603.06545 Authors: Jessica Sanson, Rahul C. Shah, Maximilian Pinaroc, Cagri Tanriover, Valerio Frascolla Categories: eess.SP, cs.AI

The Gap

Wi-Fi sensing has been stuck in a catch-22. Academic systems like WiTrack and others achieved impressive ranging accuracy, but required custom hardware modifications or specialized SDRs that nobody actually has. Meanwhile, COTS Wi-Fi sensing systems (using Intel 5300 or similar) could extract CSI from commodity cards, but were limited to coarse motion detection and presence sensing — they couldn’t tell you how far away something is with any precision. The fundamental problem: commercial Wi-Fi wasn’t designed for radar. Its limited bandwidth (80-160 MHz vs GHz for real radar) and lack of synchronized timing made accurate ranging seem impossible without hardware hacks.

Problem: COTS Wi-Fi = communication tool, not ranging sensor
   |
   v
Assumption: With proper signal processing, CSI contains range info
   |
   v
Method: Sync CSI extraction + phase alignment + self-interference cancellation
   |
   v
Evidence: Centimeter-level ranging at 40+ Hz on unmodified Intel NICs
   |
   v
Conclusion: Commodity Wi-Fi can do precision radar without hardware mods

The Increment

One sentence: Before LiveSense, you needed custom hardware to get accurate range from Wi-Fi signals; after LiveSense, your existing laptop can measure distance to centimeter precision while still browsing the web.

Core Mechanism

LiveSense operates as a three-stage pipeline running on commodity Intel Wi-Fi 6E/7 cards. First, it extracts Channel State Information (CSI) — the raw phase and amplitude measurements across all OFDM subcarriers — at a synchronized rate of 40+ Hz. This isn’t trivial: commercial Wi-Fi drivers don’t expose CSI by default, and when they do, the timestamps are unreliable. LiveSense hooks into the driver layer to grab CSI frames with precise timing.

Second, it performs time-phase alignment. Each CSI measurement comes with phase offsets introduced by the hardware’s oscillators and processing delays. These offsets are random and would destroy any ranging calculation. LiveSense tracks these offsets across frames and compensates for them, essentially calibrating the “clock” of the Wi-Fi radio on the fly. It also cancels self-interference — the massive signal from the transmitter that would otherwise drown out the tiny reflections from nearby objects.

Third, it computes Range-Doppler maps using standard radar signal processing (FFT across time for Doppler, IFFT across frequency for range). The key insight: even with just 160 MHz bandwidth, careful phase processing can extract centimeter-level range resolution. The system streams these maps, along with raw CSI and synchronized video, to a Python GUI in real-time.

   [Wi-Fi NIC]
        |
        | (CSI extraction at 40+ Hz)
        v
   [Sync Buffer] ---> timestamps aligned
        |
        | (phase calibration)
        v
   [Phase Aligner] ---> oscillator offsets removed
        |
        | (self-interference cancellation)
        v
   [Signal Cleaner] ---> TX leakage suppressed
        |
        | (FFT/IFFT processing)
        v
   [Range-Doppler Engine] ---> distance + velocity maps
        |
        v
   [GUI Stream] ---> real-time visualization

Think of LiveSense like a high-speed camera with a dirty lens. The Wi-Fi card is constantly taking “photos” (CSI snapshots) of the radio environment, but each photo has random smudges (phase offsets) and a bright spotlight in the corner (self-interference from the transmitter). LiveSense’s job is to: (1) make sure the camera shutter fires at exact intervals (synchronization), (2) figure out where the smudges are and digitally remove them (phase alignment), (3) block out the spotlight so you can see the faint objects in the background (interference cancellation), and (4) stack all the cleaned photos together to create a 3D map showing not just where things are, but how fast they’re moving (Range-Doppler processing). The camera never stops taking communication photos — it’s doing double duty.

Key Concepts

• Channel State Information (CSI): When your Wi-Fi card sends a signal, it doesn’t just blast out a single frequency — it uses OFDM, which splits the signal across dozens of narrow frequency channels (subcarriers). Each subcarrier experiences the environment differently: some get delayed by reflections off walls, others get phase-shifted by moving objects. CSI is the card’s internal measurement of “what happened to each subcarrier” — specifically, the amplitude and phase at each frequency. It’s like having 64 or 128 separate microphones listening to different pitches simultaneously. Normally this data is used internally to optimize communication (beamforming, equalization), but if you can extract it, you’ve got a detailed fingerprint of the radio environment. Example: if someone walks toward the router, the phase of higher-frequency subcarriers shifts faster than lower ones, and that differential tells you both distance and velocity.

• Self-Interference Cancellation: Your Wi-Fi card’s transmitter and receiver sit millimeters apart. When the transmitter fires, the receiver gets blasted with a signal that’s millions of times stronger than the faint echo bouncing back from a person across the room. It’s like trying to hear a whisper while standing next to a jet engine. Self-interference cancellation means estimating what the transmitter’s signal looks like at the receiver (including reflections from the laptop’s own case and nearby clutter) and subtracting it out digitally. LiveSense does this by characterizing the “static” environment when nothing is moving, then using that as a template to remove the overwhelming direct path and static reflections, leaving only the dynamic reflections from moving targets.

• Range-Doppler Map: This is a 2D grid where one axis is distance and the other is velocity. Each pixel’s brightness tells you “how much signal energy came from an object at this distance moving at this speed.” You build it by doing two Fourier transforms: one across frequency (which converts phase differences between subcarriers into distance) and one across time (which converts phase changes over successive CSI frames into velocity via Doppler shift). It’s the same math radar systems use, just applied to Wi-Fi signals. Example: if you see a bright spot at 2 meters and -0.5 m/s, there’s something moving toward you at walking speed, two meters away. Respiration shows up as a tiny oscillation at fixed distance.

Framework Shift

Before (mainstream approach):        After (this paper):

[Custom Hardware]                    [COTS Laptop]
       |                                    |
   Wideband RF                          Wi-Fi NIC
   (GHz BW)                            (160 MHz)
       |                                    |
       v                                    v
  High-res ranging                   CSI extraction
  (cm-level)                                |
       |                                    v
       X                              Phase alignment
  Can't communicate                   + interference
  simultaneously                       cancellation
                                            |
                                            v
                                      cm-level ranging
                                            |
                                            v
                                      Still communicates

From “sensing requires specialized hardware” to “sensing is a software layer on communication hardware,” the core shift is dual-use: the same radio does both jobs without compromise.

Expert Assessment

Problem choice: This is a real gap with genuine impact potential. Wi-Fi sensing has been stuck in the lab-to-product valley for a decade precisely because of the hardware barrier. If you can make commodity devices sense accurately, you unlock applications in smart homes, elderly care, and gesture interfaces without asking users to buy new hardware. The timing is good too — Wi-Fi 6E/7 adoption is accelerating, so the target hardware base is growing. Not a manufactured problem.

Method maturity: This is mostly clever engineering rather than algorithmic novelty. The core techniques (CSI extraction, phase calibration, self-interference cancellation) are well-known in the radar and SDR communities. The contribution is integration: making these techniques work together on locked-down commercial hardware with closed-source drivers. That said, the paper doesn’t explain how they achieved the driver-level CSI extraction — that’s the actual hard part, and it’s glossed over. The phase alignment approach likely uses pilot tones or known preambles, but specifics are missing.

Experimental integrity: This is a demo paper, not a full evaluation, so we don’t have baselines, accuracy distributions, or failure case analysis. The abstract claims “centimeter-level” accuracy but doesn’t quantify it (±1cm? ±5cm?). No comparison with existing COTS Wi-Fi sensing systems or ground truth from actual radar. The respiration and gesture detection claims need validation data. For a demo, this is fine, but it’s not publishable as a full paper without rigorous evaluation.

Writing quality: The abstract oversells (“first-ever demo to obtain accurate range information from commercial Wi-Fi”) — prior work has done ranging on COTS, just not at this claimed precision. The technical depth is shallow: we get a feature list but not the implementation details that would let someone reproduce this. The most valuable section would be “how we hacked the Intel driver to get synchronized CSI” — that’s completely absent. If they rewrote this with a detailed methods section and honest evaluation, it would be a strong systems paper.

Verdict: weak accept — Impressive engineering demo that solves a real deployment barrier, but lacks the rigor and detail for a full research contribution.

Takeaways

The big transferable idea: dual-use signal processing. Instead of designing systems that do one thing well, look for ways to extract secondary information from signals already being transmitted for another purpose. Communication systems are everywhere (cellular, Bluetooth, LoRa) and they all leak environmental information through their channel measurements. The pattern here — synchronize, calibrate, cancel interference, apply radar processing — is a template for turning any communication link into a sensor.

Specific technique: on-device phase alignment using static environment characterization. If you’re working with phase-sensitive measurements from hardware with unknown/drifting oscillators, you can often calibrate by measuring a known-static scenario first, then tracking deviations. This avoids needing external reference signals or hardware modifications.

Practical insight: 160 MHz is enough for centimeter ranging if you process phase carefully. The intuition that “more bandwidth = better resolution” is true for simple systems, but with coherent phase processing across subcarriers, you can beat the naive bandwidth limit. This matters for any application where you’re stuck with limited spectrum.