An Integrated Classification and Clustering System for Radio Frequency-Activity Modeling and Pattern Recognition

This effort proposes to develop an Integrated Classification and Clustering (ICC) system for AutomaticRadio Frequency Activity Modeling and Pattern Recognition (RF-AMPR). The proposed system consistsof three main components, an RF signal generative model (RFGM), a cyclostationsary feature (CF)extraction module, and the ICC system. The RFGM generates simulated RF data and captures RFactivities present in specific electromagnetic environments for emulation and testing. The CF extractionmodule constructs cyclostationsary correlation data to reveal unique RF signatures for discriminatingvarious waveforms and channel models. The main component, ICC based RF-AMPR system models andanalyzes RF activities by conducting signal classification and clustering simultaneously, and performspattern recondition and anomaly detection as required. The proposed ICC-based RF-AMPR system takesinto account “non-analyzable” discarded sensing data to gain additional insights into RF activities. Suchinsights further enable the system to detect jamming signals, monitor spectrum activities, and identifyhidden anomalies and abnormal events. To validate feasibility and scalability, a comprehensive study oneach module and the integrated system will be conducted especially under pressing scenarios with floodedRF data. Preliminary results so far showed that the proposed approach has a potential to deliver promisingperformance with scalable computation.