feasibility and limits of wi-fi imaging
TRANSCRIPT
Feasibility and Limits of Wi-Fi Imaging
Donny HuangRajalakshmi Nandakumar Shyamnath Gollakota
(University of Washington)
SenSys-2014
Goals
To explore limits of imaging using Wi-Fi signals
To describe and evaluate a prototype – Wision
Approach Overview
Use of information from Wi-Fi signal (OFDM)
Transform echo of Wi-Fi signal to image visible to humans
Extract the depth information
Line-of-sight and non-line-of-sight scenarious
Static infrastructure with several antennas
Based on prior work in domains
• Wi-Fi localization and device-free sensing (MUSIC etc.)• Motion detection• RF Image Sensing (e.g. radar systems)• Radio Tomography Imaging• Magnetic Resonance Imaging (MRI)• Optics• Microscopy• Computational Photography
Advantages of Wi-Fi over Radar Imaging
Cost of Wi-Fi chipset is lower than a radar device
Bi-static systems have wider range of operation (for the same power level)
Wi-Fi devices needed for “lighting” are already there
Challenges using Wi-Fi for Imaging
Wi-Fi offer much lower bandwidth than is required for radar imaging method (20MHz -> 7.5m resolution)
OFDM-based transmissions are not designed for imaging purposes (OFDM – Orthogonal Frequency Division Multiplexing)
Wision Design
Different approach: Measuring the incoming signal at each azimuth and elevation angle – like in optical imaging systems
Signal is a linear combination of reflections from multiple regions on each antenna
Separation of signals from different directions is done by – multiple antennas and Fourier analysis
Wision’s Imaging Algorithm
Algorithm summary:
1. Initialization. The receiver measures phase and magnitude information from all of its antennas.
2. Step 1. Compute the azimuthal (ψ) and elevation (α) angles for a given region in space.
3. Step 2. Compute the corresponding intensity value by performing 2D FFT.
4. Step 3. Repeat Step 1 and Step 2 for every region in the 2D-space to generate image visible by human eye.
Accounting for the Wi-Fi Transmitter
• Removing direct strong Wi-Fi transmissions using multiple antennas and interference nulling
• Using multiple transmitters
Wi-Fi Interaction with Objects
Factors:
• Material of the object• Size of the object (~ to the wavelength, 12cm at 2.4GHz)• Diffraction effects
Implementation
Wision prototype:
• USRP-N210 hardware (Unified Software Radio Peripheral)• 2.4 GHz frequency• 10 MHz bandwidth• 8x8, 4x4, 8x1, 4x1 antenna arrays• Linear actuator (8 x 8, 6cm interval)• wa5vjb or hg2415g directional antenna, 0..180o, by step
10o (for beamforming tests)
Imaging Using 1-D Antenna Arrays
Objects at different receiver directions
2 metallic objects (16x20x5cm)
8x1 antenna array at receiver
Imaging Using 1-D Antenna Arrays
Objects at different receiver depths. Experiment 1
2 metallic objects (44x44x18cm) at 80o(2m) and 130o(0.8m) from the receiver
8x1 antenna array at receiver, directional antenna wa5vjb at transmitter
Imaging Using 1-D Antenna Arrays
Objects at different receiver depths. Experiment 2
2 objects – a stack of books and a laptop one behind another (at 1 and 3 meters)
8x1 antenna array at receiver, hg2415g directional antenna at transmitter
Imaging Using 1-D Antenna Arrays
Objects at different receiver depths. Observed limitations
1. Orientation of objects is a key factor (light must reach the receiver)
2. If objects of different materials are used, objects with material with strong reflection will dominate in a image
Imaging Using 2-D Antenna Arrays
T-shaped metallic object8x8 antenna array as a receiver, 1-2 antennas as a transmitter
Imaging Using 2-D Antenna Arrays
Observed limitations
1. 10 minutes per image with 8x8 antenna array (moving antenna by hand)
2. Objects with material with strong reflection will dominate in a image
Imaging in NLOS Scenarios
Observed limitations
1. As the barrier would become thicker and from more reflective material – the less transparent it would be
2. In the experiment setting coach was very near to wall. Its mainly because OFDM implementation on USRP’s did not work good with high power levels (due to hw non-linearities)
Proof-of-Concept Applications
1. Localizing Objects without Tagging Them
• Room dimensions 35m2 with furniture (coach, tables, chairs)• Object to localize - desktop 44x44x18cm• Room does not have other similar objects• At 10 different locations:• Imaging using 2 antenna array receivers• Combine• Search for a high reflection area (accuracy 14cm)• Localization error computed
• Beamforming and nulling algorithms
Proof-of-Concept Applications
2. Static Human Localization
• Human, male, 174cm, 65kg• Setting as above (but no furniture)• 10 different locations (static)• Other humans moved around
Proof-of-Concept Applications
4. Imaging Through-the-fabric Scenarios. Detecting phone in the pocket
The Limits of Wi-Fi Imaging
Object Size and Material
Imaging Resolution
Orientation of Objects
Questions?