Cognitive Engine Design for Spectrum Situational Awareness and Signals Intelligence
Sudharman K. Jayaweera, Mohamed A. Aref · 2018
Cognitive radio technology is proposed as a means to achieve real-time spectrum situational awareness (SSA) and signals intelligence (SIGNIT) over a wide spectrum range by designing a cognitive engine that performs machine-learning based hierarchical RF signal identification. This allows the radio to classify and associate a signal using as fewer a number of features as possible. The classification algorithms can be based on any suitably chosen machine-learning algorithm such as artificial neural networks (ANNs) or deep learning. The proposed design allows the user to define, and modify, the SSA parameters during field operations. The specific example design proposed in this paper allows these definitions to be based on two levels of signal classification: a broad type of signals such as communications or radar and specific signals within each of these classes. It is shown that not only the proposed cognitive engine design allows realizing a large number of SSA definitions using permutations of the same two stage classifiers, but also the hierarchical approach may outperform dedicated classifiers with similar computational complexity. The design can easily be generalized to handle more than two levels of signal classification.