Detection of traffic patterns in the radio spectrum for cognitive wireless network management

摘要: Dynamic Spectrum Access allows using the spectrum opportunistically by identifying wireless technologies sharing same medium. However, detecting a given technology is, most of time, not enough to increase efficiency and mitigate coexistence problems due radio interference. As solution, recognizing traffic patterns may lead select best time access shared optimally. To this extent, we present recognition approach that, our knowledge, is first non-intrusive method detect directly from spectrum, contrary traditional packet-based analysis methods. In particular, designed Deep Learning (DL) architecture that differentiates between Transmission Control Protocol (TCP) User Datagram (UDP) traffic, burst with different duty cycles, varying rates transmission. input these models, explore use images representing in time-frequency. Furthermore, novel data randomization generate realistic synthetic combines two state-of-the-art simulators. Finally, show after training testing models generated dataset, achieve an accuracy ≥ 96 % outperform methods based on IP-packets DL.