Better distribution of cognitive capacities is one way to enable the integration of more functions into sensors and eliminate the need to place management capabilities in the cloud. However, more functions means systems that are more complex to build and that consume more energy. Compressed data acquisition compensates for these disadvantages by acquiring and processing only the most relevant information. Researchers at Leti, a CEA Tech institute, tested the approach on radiofrequency signals, analyzing the spectral occupancy of ambient signals.

The approach could be useful for radio receivers and cognitive radar. A radiofrequency signal never occupies all of its emission spectrum—this is known as signal parsimony. Detecting which frequency bands are actually occupied is the first step toward focusing radio reception on occupied bands only and, more broadly, adapting the receiver to its environment.

Leti researchers recently developed and patented a compressed acquisition architecture in conjunction with Cornell University (United States). The architecture was integrated into an ASIC developed in-house and was proven effective up to a degree of parsimony of 12% (the useful signal takes up a maximum of 12% of the spectrum). It uses ten times less energy than the current state of the art. A PhD research project on radar applications and a future EU project to further improve the results obtained will build on Leti's work.