PaperlayerView exact revision on arXiv

Source-linked AI summary

FarSense: Pushing the Range Limit of WiFi-based Respiration Sensing with CSI Ratio of Two Antennas

Youwei Zeng, Dan Wu, Jie Xiong, Enze Yi, Ruiyang Gao, Daqing Zhang

arXiv:1907.03994v3eess.SPcs.HC

TL;DR

Existing WiFi respiration-sensing systems have limited range, restricting use when targets are far from transceivers or separated by a wall. FarSense uses a two-antenna CSI ratio to reduce noise, recover stable phase information, and combine amplitude and phase. The system extends sensing from 2-4 meters to 8-9 meters and enables through-wall respiration sensing with commodity WiFi hardware.

  • Problem

    Existing WiFi respiration-sensing approaches work best near the transceivers, with sensing limited to 2-4 meters despite WiFi communication ranges of tens of meters.

  • Method

    FarSense divides CSI readings from two antennas to cancel most noise and time-varying phase offset, then combines the ratio’s amplitude and phase for respiration sensing.

  • Results

    The sensing range increases from 2-4 meters to 8-9 meters, and FarSense enables through-wall respiration sensing with commodity WiFi hardware.

  • Takeaways & Limitations

    FarSense moves WiFi-based respiration sensing toward house-level, real-life deployment and may benefit other sensing applications.

  • Takeaways & Limitations

    The method assumes one dominating reflection path from the human chest for respiration sensing.

Abstract

from arXiv · show

The past few years have witnessed the great potential of exploiting channel state information retrieved from commodity WiFi devices for respiration monitoring. However, existing approaches only work when the target is close to the WiFi transceivers and the performance degrades significantly when the target is far away. On the other hand, most home environments only have one WiFi access point and it may not be located in the same room as the target. This sensing range constraint greatly limits the application of the proposed approaches in real life. This paper presents FarSense--the first real-time system that can reliably monitor human respiration when the target is far away from the WiFi transceiver pair. FarSense works well even when one of the transceivers is located in another room, moving a big step towards real-life deployment. We propose two novel schemes to achieve this goal: (1) Instead of applying the raw CSI readings of individual antenna for sensing, we employ the ratio of CSI readings from two antennas, whose noise is mostly canceled out by the division operation to significantly increase the sensing range; (2) The division operation further enables us to utilize the phase information which is not usable with one single antenna for sensing. The orthogonal amplitude and phase are elaborately combined to address the "blind spots" issue and further increase the sensing range. Extensive experiments show that FarSense is able to accurately monitor human respiration even when the target is 8 meters away from the transceiver pair, increasing the sensing range by more than 100%. We believe this is the first system to enable through-wall respiration sensing with commodity WiFi devices and the proposed method could also benefit other sensing applications.

1 INTRODUCTION

FarSense addresses the limited range of WiFi-based respiration sensing by using CSI ratios from two antennas and combining amplitude with phase information. The system targets house-level sensing and challenging real-world deployments, including through-wall scenarios.

  • Approach: FarSense forms a CSI ratio from two antennas instead of using raw single-antenna CSI readings.The division operation cancels most original CSI noise and time-varying phase offset, producing a more sensitive signal for subtle movements.
  • Evaluation: FarSense evaluates subjects at different distances, across a wall between transceivers, and with both transceivers mounted on the ceiling.These scenarios represent challenging placements for real-world deployment.
  • Approach: The system combines CSI-ratio amplitude and phase information to address blind spots and further extend respiration-sensing range.The ratio makes phase information stable enough to use alongside amplitude.
  • Results: The sensing range increases from the current state-of-the-art 2-4 meters to 8-9 meters, enabling through-wall respiration sensing with commodity WiFi hardware.The authors present this as a step toward real-life deployment.

2 RELATED WORK

Prior WiFi respiration systems primarily use CSI amplitude or attempt to calibrate noisy phase information, while radar approaches use specialized sensing technologies. FarSense instead uses a two-antenna CSI ratio whose amplitude and stable phase are combined for respiration sensing.

  • Radar-based respiration sensing: Radar-based respiration sensing includes continuous-wave Doppler, ultra-wideband pulse, and FMCW radar approaches.These approaches differ in radio architecture, bandwidth, multipath handling, and distance measurement capabilities.
  • WiFi-based respiration sensing: CSI amplitude contains relatively large noise from power amplifier uncertainty and environmental noise, despite its mathematical and physical relationship to human movement.Amplitude-based systems also retain the blind-spots issue.
  • WiFi-based respiration sensing: Existing WiFi respiration systems mainly use CSI amplitude, while phase-based approaches face noise from SFO, CFO, and PDD.Reported phase-calibration methods still fail to monitor fine-grained millimeter-level chest motion caused by respiration.
  • FarSense: FarSense introduces a two-antenna CSI ratio that cancels most amplitude noise and time-varying phase offset, enabling stable phase information for sensing.It combines ratio amplitude and phase because they provide complementary sensing capabilities and help remove blind spots.

3 EMPIRICAL STUDY

The empirical study compares CSI amplitude ratios with raw single-antenna amplitudes using a moving metal plate and examines the ratio’s amplitude and phase properties. The amplitude ratio produces clearer movement-induced signal variations, particularly as the target moves farther from the transceivers.

  • Experimental Settings: The study moves a 15 cm square metal plate along the transceiver pair’s perpendicular bisector from 5 m to 7 m in an empty room.The transceivers and plate are positioned at the same height, with a 5.5 m line-of-sight path.
  • Experimental Results: The experiment compares two raw CSI amplitude waveforms with the amplitude ratio from two antennas as the plate moves farther away.The ratio’s amplitude equals the ratio of the two raw CSI amplitudes.
  • Experimental Results: The plate movement pattern is buried in raw amplitude noise but appears much clearer in the two-antenna amplitude ratio.The division operation cancels most raw-amplitude noise, including high-amplitude impulses and burst noise.
  • Experimental Results: When the line-of-sight path length varies, the amplitude ratio continues to show clearer movement-induced variations, especially when the target is farther away.This observation is reported across the varied path-length settings.
  • CSI Ratio Properties: The CSI ratio remains a complex number whose amplitude is an amplitude ratio and whose phase is the phase difference between the two antennas.The study therefore explores both amplitude and phase properties of the ratio for sensing.

4 THE CSI-RATIO MODEL

The CSI-ratio model represents indoor WiFi propagation through static and dynamic components, then uses the ratio between two antennas to model movement-related changes geometrically. Its properties are verified through benchmark experiments showing circular trajectories, orientation changes, and path-length-dependent arcs.

  • CSI Primer: CSI is modeled as a superposition of signals from direct and reflected propagation paths.The paths are grouped into a static component and a dynamic component reflected from the moving human target.
  • CSI Primer: A short movement keeps the dynamic-component amplitude approximately constant, while changing path length rotates CSI in the complex plane.A path-length increase of one wavelength produces a 2π rotation; smaller changes produce circular arcs.
  • CSI Ratio: The CSI ratio divides measurements from two antennas, relying on a shared phase offset and approximately constant inter-antenna path-length difference.The resulting expression has the form of a Möbius transformation and can be decomposed into translations, complex inversion, and complex multiplication.
  • CSI Ratio: Complex inversion preserves circle shape but reverses its rotation orientation when the circle contains the origin.Translations and complex multiplication alter position, scale, or rotation but do not change the circle's geometric shape or orientation in the same way.
  • Model Verification: Benchmark experiments confirm that CSI-ratio trajectories form circles, switch clockwise or counterclockwise orientation with static-to-dynamic magnitude, and trace arcs proportional to path-length changes.A one-wavelength change yields a full circle and a sub-wavelength change yields a corresponding circular arc; similar behavior appears at 5.75 meters.
  • Model Verification: The validated CSI-ratio properties are presented as applicable beyond respiration sensing, including tracking, motion detection, fine-grained movement, and other wireless technologies.The paper focuses on using the model to extend WiFi respiration-sensing range.

5 EXTRACTING RESPIRATION PATTERN FROM CSI RATIO

FarSense extracts respiration from CSI ratios by combining amplitude and phase information through projections in the complex plane. It generates many projection candidates and selects the most periodic one, improving extraction when distant targets make the ratio noisy.

  • FarSense combines the amplitude and phase of CSI ratios to extract respiration patterns from distant targets and extend sensing range.
  • CSI-ratio phase remains usable because division cancels the time-varying random phase offset, while the I and Q components provide complementary sensing information.
  • 5.2.1 Generating Combination Candidates: Projection maps complex CSI-ratio data onto axes whose weights are cosθ for I and sinθ for Q, generating fine-grained I/Q combinations.
  • 5.2.1 Generating Combination Candidates: 2n projection candidates are generated by varying θ from 0 to 2π in steps of π/2^n, whereas the traditional I/Q selection provides only two candidates.
  • 5.2.2 Selection from Multiple Candidates: The candidate with maximal short-term breathing-to-noise ratio is selected, using respiration energy relative to total energy as the periodicity criterion.
  • 5.2.2 Selection from Multiple Candidates: When the target is far away, maximal-periodicity projection reveals a clear respiration pattern while amplitude, phase, and maximal-variance alternatives do not.

6 THE FARSENSE SYSTEM

The FarSense system collects and preprocesses two-antenna CSI, extracts respiration patterns across sub-carriers, and estimates respiration rate by combining autocorrelation results. It excludes unstable periods and low-quality sub-carriers to support real-time operation.

  • FarSense comprises Data Collection, Data Preprocessing, Respiration Pattern Extraction, and Respiration Rate Estimation modules.
  • 6.2 Data Preprocessing: The system divides complex CSI readings from two receiver antennas for each sub-carrier to obtain a complex CSI ratio.
  • 6.2 Data Preprocessing: Periods containing large human motions are excluded because minute chest movements are overwhelmed during high mobility, and stable periods are smoothed with a Savitzky-Golay filter.
  • 6.3 Respiration Pattern Extraction: 100 projection candidates are generated for each of 30 sub-carriers, and the candidate maximizing BNR is selected; processing all sub-carriers takes about 0.54 s.
  • 6.4 Respiration Rate Estimation: Autocorrelation is applied to each sub-carrier’s respiration pattern, then the results are combined using sub-carrier BNR values as weights.
  • 6.4.2 Multiple Sub-carriers Combining: Sub-carriers with BNR below 0.7ε are excluded because distant sensing occasionally produces chaotic patterns from differing multipath and shadowing effects.
  • 6.4.2 Multiple Sub-carriers Combining: Using three selected sub-carriers, the system estimates 17.9 bpm from a first autocorrelation peak at lag 335 with a 100 Hz sampling rate.

7 EVALUATION

The evaluation uses comprehensive experiments to assess FarSense against state-of-the-art methods, including sensing range and a scenario where the access point is in another room.

  • The evaluation compares FarSense with state-of-the-art approaches for sensing range and tests robustness when the WiFi access point is in another room.

7.1 Experimental Setup

Experiments use commodity Intel 5300 WiFi hardware and real-time MATLAB processing, with a graphical interface showing respiration and human-status information.

  • The setup uses two GIGABYTE mini-PC transceivers with Intel 5300 WiFi cards, one transmitter antenna, and two receiver antennas at 5.24 GHz.
  • The threshold for excluding sub-carriers is empirically set to 0.7 after testing different values with large amounts of data.
  • CSI is processed in real time with MATLAB on a DELL Precision 5520 laptop, while the GUI displays respiration pattern, test-environment video, human status, and estimated respiration information.

7.2 Comparison with Previous Approaches

FarSense is compared with HRD and FullBreathe using respiration-pattern clarity and detection rate across increasing target distances. It maintains reliable sensing substantially farther than both prior approaches.

  • Respiration Pattern: At 5 m, FarSense retains clear respiration patterns, whereas HRD and FullBreathe fail to extract them.At 4 m, HRD is noisy while FullBreathe and FarSense remain clear.
  • Overall Detection Rate: 100% detection rate at 5 m is achieved by FarSense, compared with 1.4% for HRD and 7.7% for FullBreathe.The corresponding sensing ranges are less than 2.9 m, 3.7 m, and larger than 5.7 m, respectively.
  • Overall Detection Rate: FarSense increases sensing range beyond the limits observed for HRD and FullBreathe without sacrificing accuracy.The comparison uses the defined threshold of detection rate higher than 95%.

7.3 Performance in Challenging Real-life Scenarios

FarSense is evaluated beyond 5 m, through-wall, and in ceiling-mounted sleeping scenarios. It continues to monitor respiration with low error across these challenging real-life settings.

  • Sitting Far from the Transceivers: 0.28 bpm mean absolute error is observed at 6 m, increasing to 0.64 bpm at 9 m.Mean absolute error increases with distance as the reflected signal is further attenuated.
  • Sitting Far from the Transceivers: At 8 m, FarSense achieves 100% detection with mean absolute error below 0.5 bpm.Its sensing range is therefore larger than 8 m, versus less than 2.9 m for HRD and 3.7 m for FullBreathe.
  • NLoS Scenarios: 0.34 bpm mean absolute error is achieved across the evaluated NLoS scenarios with a wall between transmitter and receiver.The other two state-of-the-art systems fail in these NLoS scenarios.
  • Sleeping Scenarios: Below 0.3 bpm mean absolute error is achieved across four sleeping postures with ceiling-mounted transceivers and a quilt covering the subject.Supine posture is more accurate than the other postures.

8 DISCUSSIONS

The discussion identifies scope boundaries for FarSense and directions for extending it to harder deployment conditions. These include double-wall paths, multiple subjects, wider bandwidth, and more antennas.

  • 8.1 Through the Wall Twice: FarSense has difficulty sensing respiration when both transceivers are in another room because the reflected chest signal crosses the wall twice.The paper suggests nulling the strong static LoS signal and other reflections as a potential solution.
  • 8.2 Multiple Subjects: Multiple-subject respiration sensing remains challenging with cheap commodity hardware.The paper identifies increased bandwidth or more antennas as possible ways to separate multiple reflected signals.
  • Future Directions: The proposed CSI-ratio model can be combined with multiple-antenna techniques to increase sensing range and resolution.MUSIC and multidimensional information fusion are suggested for further signal separation and processing.
  • System Scope: FarSense uses the ratio of CSI readings from two receiving antennas and does not depend on a specific MIMO mechanism.The CSI streams must share the same clock, as in current commodity MIMO WiFi NICs.

9 CONCLUSION

FarSense uses CSI ratios and complementary amplitude-phase information to extend contactless respiration sensing from room level to house level. The paper reports through-wall sensing with commodity WiFi hardware.

  • Conclusion: FarSense pushes respiration sensing from 2-4 m to 8-9 m using ratios of CSI readings from two adjacent antennas.The ratio is presented as a new base signal for MIMO devices and other sensing applications.
  • Conclusion: Complementary amplitude and phase information is combined to improve sensing accuracy and range.The approach addresses blind spots in respiration sensing.
  • Conclusion: FarSense enables through-wall respiration sensing with commodity WiFi hardware.The conclusion frames this as reducing the gap between lab prototypes and real-life deployment.
Loading 1907.03994v3…