A Software Defined Radio Testbed for Over-the-air Cognitive Cycle Demonstration

Abstract

Deep learning (DL) have been widely applied in cognitive radio, including cognitive jamming, cognitive communication and cognitive radar. Many functional properties of DL are amenable to numerous electromagnetic waveform recognition tasks. In fact, there exists a gap between the DL network design and the real time application. This prompts us to adopt the software defined radio testbed to realize the online "cognition-action" demonstration. Via an innovative artificial intelligent (AI)-baseband co-design, the system can realize the modulation recognition and demodulation adaption, which is associating with a demonstration of "cognition-action" cycle. In addition, to realize the online recognition and adaption, we design the over-the-air demodulation reconstruction method. By our experimental results, we demonstrates that such cognitive cycle can bring about noticable improvement in cognitive applications.