Deep Cognitive Radio System Study

Abstract

This study investigates the applicability of feature learning and deep generative representation of radio signals drawing from state of the art machine learning techniques especially in automatic voice recognition, image representation and machine translation domains. It seeks to demonstrate the feasibility of naively learning intelligent radio transponder baseband algorithm behavior without significant prior expert knowledge about signal structure and signal processing algorithms. This study seeks to outline what a future radio system architecture applying these principals might look like, how it might be structures, which classes of algorithms it may focus on, and what technical requirements it may have. By investigating this basic approach to learned radio behavior we seek to improve significant problems in radio interoperability, adaptation, and protocol fragmentation which currently plague countless government and commercial radio systems making safe, efficient and adaptive interoperability difficult.

Document Details

Document Type
DoD Grant Award
Publication Date
Feb 11, 2016
Source ID
HR00111610002

Entities

People

  • Timothy O Shea

Organizations

  • Defense Advanced Research Projects Agency
  • Virginia Tech

Tags

Fields of Study

  • Computer science

Readers

  • Neural Network Machine Learning.
  • Tactical Satellite Communications Systems Engineering.

Technology Areas

  • AI & ML
  • AI & ML - Neural Networks