The explosion of 5G networks and the Internet of Things is expected to result in an exceptionally crowded RF environment, where techniques such as spectrum sharing and dynamic spectrum access will be essential components of the wireless communication process. In this vision, wireless devices must be able to learn to autonomously extract knowledge from the spectrum on-the-fly; and react in real time to the inferred spectrum knowledge by appropriately changing communication parameters, including frequency band, symbol modulation, coding rate, among others. Traditional CPU-based machine learning suffers from high latency, and requires application-specific and computationally-intensive feature extraction/selection algorithms. 

Conversely, deep learning allows the analysis of massive amount of unprocessed spectrum data without ad-hoc feature extraction. So far, deep learning has been used for offline wireless spectrum analysis only. Therefore, additional research is needed to design systems that bring deep learning algorithms directly on the device’s hardware and tightly intertwine with the RF components to enable real-time spectrum-driven decision-making at the physical layer. 

Technology Overview

Northeastern University inventors present RFLearn, the first system enabling spectrum knowledge extraction from unprocessed in-phase quadrature (I/Q) samples by deep learning directly in the RF loop. RFLearn provides a complete hardware/software architecture where the CPU, radio transceiver and learning/actuation circuits are tightly connected for maximum performance; and a learning circuit design framework where the latency vs. hardware resource consumption trade-off can be explored. 

A custom software-defined radio built on a system-on-chip (SoC), mounting radio transceivers and antennas was implemented. This was applied to solving the fundamental problems of modulation and OFDM parameter recognition. Experimental results reveal that RFLearn decreases latency and power by about 17x and 15x with respect to a software-based solution, with a comparatively low hardware resource consumption.


  • Enables for the first time spectrum-driven real-time decision-making without CPU involvement
  • Effective and efficient hardware design
  • Reduce latency by 15x
  • Reduce power by 17x
  • Fully reconfigurable through software
  • Can optimize latency/space/energy tradeoff


Wide range of military and civilian applications, including:

  • Spectrum sensing for (i) detection of adversarial action (e.g., jamming) and (ii) system throughput optimization through dynamic tuning of critical RF parameters
  • Smart dynamic spectrum management for increased spectrum efficiency (e.g., through beam forming)


  • Development partner
  • Commercial partner
  • Licensing


  • Development partner
  • Commercial partner
  • Licensing

IP Status

  • Patent application submitted


Patent Information:
For Information, Contact:
Mark Saulich
Associate Director of Commercialization
Northeastern University
Tommaso Melodia