Seeing Through Walls of Unknown Materials

A sparse blind deconvolution method for imaging through layered media.

A view of a microwave scan of the Duke logo taken through a wall before and after distortions have been removed. By taking into account the types of distortions typically created by flat, uniform walls, the new algorithm allows for better scans without needing to know what the wall is made of beforehand.

January 29, 2018 | Source: Duke University, pratt.duke.edu, 6 Dec 2017, Ken Kingery

Researchers at Duke University have devised a way to see through walls using a narrow band of microwave frequencies without any advance knowledge of what the walls are made out of. Besides having obvious applications in the realm of security, the approach could lead to inexpensive devices to help construction workers easily locate conduits, pipes and wires.

“Most technologies that can see through walls use a broad range of frequencies, which makes them expensive,” said Daniel Marks, associate research professor of electrical and computer engineering at Duke. “They also don’t have very good resolution. So while they might be fine for seeing a person moving on the other side of a wall, they’re terrible for finding thin conduits or wires.”

Current approaches also typically rely on knowing what material the wall is made out of before trying to see through it. This allows the software to predict how the wall will affect the scanning waves so that it can separate the echoes and distortions from the solid objects being sought.

In the new paper, Marks and his colleagues David R. Smith, the James B. Duke Professor of Electrical and Computer Engineering, and Okan Yurduseven, a postdoctoral researcher in electrical and computer engineering at Duke, take advantage of a wall’s symmetry instead.

Because walls are generally flat and uniform in all directions, they distort waves in a symmetrical fashion. The newly described technology uses this symmetry to its advantage.

“We wrote an algorithm that separates the data into parts—one that shows circular symmetry and another that doesn’t,” explained Yurduseven. “The data that doesn’t have any symmetry is what we’re trying to see.”