Researchers in UC Santa Barbara Professor Yasamin Mostofi’s lab have proposed a new foundation that can enable high-quality imaging of still objects with only Wi-Fi signals. Their method uses the Geometrical Theory of Diffraction and the corresponding Keller cones to trace edges of the objects.
The technique — which appeared in the Proceedings of the 2023 IEEE National Conference on Radar (RadarConf) — has also enabled, for the first time, imaging, or reading, the English alphabet through walls with Wi-Fi, a task once deemed too difficult for Wi-Fi due to the complex details of the letters.
“Imaging still scenery with Wi-Fi is considerably challenging due to the lack of motion,” said Mostofi. “We have then taken a completely different approach to tackle this challenging problem by focusing on tracing the edges of the objects instead.”
This innovation builds on previous work in the Mostofi Lab, which since 2009 has pioneered sensing with everyday radio frequency signals such as WiFi for several different applications, including crowd analytics, person identification, smart health, and smart spaces.
“The core fundamental idea is focusing on tracing the edges of the objects, by using the Geometrical Theory of Diffraction (GTD),” Mostofi told Tech Briefs in an exclusive interview, the entirety of which can be read below. “When a given wave is incident on an edge point, a cone of outgoing rays emerges according to the GTD, referred to as a Keller cone. For a given incident wave, the cone’s shape and orientation change based on the orientation of the edge.
“Thus, depending on the orientation of the edge, it will leave different footprints, formally known as conic sections, on a given receiver grid nearby. We have then developed a new mathematical framework that uses these conic footprints as signatures to infer the orientation of the edges, thus creating an edge map of the scene.”
The team has also extensively studied the impact of several different parameters, such as curvature of a surface, edge orientation, distance to the receiver grid, and transmitter location on the Keller cones and their proposed edge-based imaging system, thereby developing a foundation for a methodical imaging system design.
In the team’s experiments, three off-the-shelf Wi-Fi transmitters send wireless waves in the area. Wi-Fi receivers are then mounted on an unmanned vehicle that emulates a Wi-Fi receiver grid as it moves. The receiver measures the received signal power which it then uses for imaging, based on the proposed methodology.
The researchers have extensively tested this technology with several experiments in three different areas, including through-wall scenarios. In one example application, they developed a Wi-Fi Reader to showcase the capabilities of the proposed pipeline.
Overall, the proposed approach can open up new directions for RF imaging.
Here is the Tech Briefs interview — edited for length and clarity — with Mostofi.
Tech Briefs: I’m sure there were too many to count, but what was the biggest technical challenge you faced while developing this imaging system?
Mostofi: That is a great question. My lab has been working on sensing with Wi-Fi signals since 2009, with our first imaging result with Wi-Fi officially published in a 2010 paper. Since then, we have enabled many different applications with Wi-Fi from imaging still objects, and crowd analytics, to smart health, and person identification.
However, while we have shown good progress over the years, imaging still objects with Wi-Fi has remained the most challenging, due to the lack of motion, as compared to the other applications. This was then the main motivation for this paper, in which we took a completely different approach to this problem by focusing on the edges of the objects.
Tech Briefs: What’s the next step? Do you have any plans for further research?
Mostofi: Our next step for this research is to enable imaging of even more complex scenes with Wi-Fi.
We have focused on imaging objects that have intricate details to showcase the potentials of the proposed approach. This was the main motivation for imaging the English alphabet as they have complex details. In addition, we have imaged a number of household items that have intricate details.
Tech Briefs: What kind of new directions could this lead to?
Mostofi: Imaging still objects is important for many RF sensing applications as it is key to context inference and scene understanding. In addition, it can find applications in smart home, smart spaces, structural health monitoring, search and rescue, surveillance, and excavation domains.
Tech Briefs: Do you have any advice for researchers aiming to bring their ideas to fruition?
Mostofi: Getting an idea to fruition not only requires many hours of hard work but will also involve not getting discouraged by the setbacks that one would naturally face. That is what I tell my lab members: to be bold, keep at it, do the hard work, and most importantly enjoy the process.
Tech Briefs: Is there anything else you’d like to add?
Mostofi: I would like to acknowledge my Ph.D. student Anurag Pallaprolu as well as now-graduated Ph.D. student Belal Korany, who have worked very hard on this project.
Transcript
00:00:00 foreign thing with Wi-Fi signals has gained considerable Traction in recent years with many potential applications Imaging still objects with RF signals however is still a considerably challenging problem due to the lack of motion in this research we have proposed a completely different approach that
00:00:24 enables Wi-Fi signals to image still objects our approach is based on the geometrical theory of diffraction and by exploiting the corresponding color cones it is worth noting that this work is not based on collecting RF data to train a neural network consider the C if we are asked to represent this scene we would probably produce something like
00:00:45 this in other words we would trace the edges tracing the edges is a compressive yet informative representation of the scene in this work we introduce y-fract which can generate high quality images of objects via Edge Tracy why fract is based on the phenomenon of H diffraction and exploits signatures that edges leave on the receiver grid it can even enable
00:01:08 Wi-Fi to read through walls as we shall see now let's see how an edge Point interacts with the incoming wave when a wave is incident on an edge a cone of outgoing Rays emerges according to Keller's geometrical theory of diffraction this interaction is not limited to visibly sharp edges but applies to a
00:01:27 broader set of surfaces with a small enough curvature next consider a receiver Grid in the vicinity of the edge depending on the orientation of the edge and the direction of the incident Ray the edges leave different signatures on the receiver grid we then propose to exploit these signatures for Imaging the edges
00:01:47 let's see how we do it here is an approximated expression for the received CSI signal power we then propose the following Edge based backward rate facing Imaging kernel more specifically by using the corresponding conic section as the signature of an edge point we find the edge orientation that maximizes the Imaging kernel via hypothesis testing
00:02:10 edges of real life objects have local dependencies we then next model the Imaging space as a graph and propagate the information of the inferred high confidence edges through the graph using Bayesian information propagation finally we can further improve the final image by using existing image completion tools from the area of vision let's see some experimental results
00:02:33 we first consider to sample non-through wall areas we have three Wi-Fi transmitters an unmanned vehicle then emulates an antenna array with a Wi-Fi receiver antenna whose CSI power measurements are then used for Imaging the next show how Wi-Fi can read with our proposed approach here are some sample letters and how they are imaged
00:02:56 see the ground truth image as well as our Imaging result as can be seen our approach can image the details of the letters pretty well with only Wi-Fi signals here is a summary of all the image letters i can be seen the letters are imaged well with only Wi-Fi CSI measurements showing the first demonstration of Wi-Fi reading
00:03:31 next let's consider Imaging through walls here is a through wall Wi-Fi Imaging setting where we are interested in reading the letters behind the wall here are the letters behind the wall and here is how our system image them with Wi-Fi as can be seen the proposed approach enables Wi-Fi to read through walls for the first time see the paper for more details and
00:03:54 results
