• Login
  • Register

Work for a Member company and need a Member Portal account? Register here with your company email address.

Project

Reconstructing Hidden Objects with Wireless Signals

Copyright

Signal Kinetics

Signal Kinetics 

What if devices could see through walls, boxes, and clutter?

We’ve developed mmNorm, a new technology that creates 3D models of objects—even when they’re completely hidden from view. While traditional cameras and LiDAR systems can only detect and recontruct what’s in their direct line of sight, mmNorm uses millimeter-wave (mmWave) radar—the same kind of wireless signal used in 5G networks and airport scanners—that can pass through common materials like cardboard, fabric, and plastic. 

This technology could enable robots to find and pick up items inside closed containers, allow AR headsets to reveal objects behind furniture, and help smart devices understand gestures even when users are out of sight.

How does it work?

Instead of simply measuring the strength of radar reflections (as past methods do), mmNorm estimates the curvature of hidden objects by analyzing how radar waves bounce off them. This allows it to reconstruct the object's shape with much greater accuracy.

Here’s the process:

  • Estimate Surface Normals
    mmNorm determines which direction each part of the hidden object surface is facing, based on the patterns of radar reflections.
  • Reconstruct the Surface Candidates
    It then pieces together these surface directions to form multiple surface candidates for the object's shape.
  • Optimize the Result
    Finally, mmNorm simulates how different 3D shapes candidate would reflect radar signals and selects the one that best matches the actual radar measurements.

We tested mmNorm on over 60 everyday objects—including mugs, tools, and toys—hidden behind boxes and clutter. You can see examples of these reconstructions in our video:

Implementation

To build mmNorm, we developed a full end-to-end prototype using commercially available hardware. We mounted a Texas Instruments  mmWave radar on a UR5e robotic arm, which moves the radar across a 2D grid to capture measurements from multiple angles.

You can read the paper here: