← All Research

Semantic & Compressive sensing

Future networks need to sense their surroundings and extract the information that actually matters. We build sensing pipelines on metasurfaces and phased arrays that capture task-relevant features, not every raw measurement.

Semantic & Compressive sensing
From left to right: scene with semantic components --> measurement via analog sensor (RIS/metasurface) --> latent compressive representation --> semantic extraction.

Future wireless and sensing systems must be aware of their surroundings — to adapt to changing channels and to extract the information that matters for imaging or localization. We develop sensing pipelines on programmable metasurfaces and phased arrays that acquire only the task-relevant features of a scene, rather than every raw measurement.

The approach grows out of the semantic signal processing framework that we helped establish, together with our work on compressive sensing for direction finding — sparse arrays, adaptive measurement-matrix design, and gridless estimation.

  • Scene reconstruction from minimal RF measurements
  • Compressive channel acquisition with sparse arrays
  • RIS hardware testbeds