A Deep Neural Network Based Environment Sensing in the Presence of Jammers

In this paper, we introduce the concept of environment sensing, which is one of the essential requirements for next-generation radio networks to enable highly dynamic radio resource allocation among multiple services. Such requirements motivate the investigation of strategies for transmitter detection, assuming that multiple sensing units (SUs) cooperate to sense the radio environment. A grid-based deployment of the SUs together with randomly placed transmitter units (TUs) and jammer units (JUs) moving with a constant speed of 3 km/h while maintaining the minimum distance between them by 3 m are considered. Active transmitter detection is done using deep neural network. Transmitter detection is performed in two steps, first, jammer detection is performed, and then, legit transmitter detection is performed. Moreover, two well-known performance metrics are employed for verifying the scheme’s behavior. The computation of the receiver operating characteristic curve and the probability of detection are used to evaluate the transmitter detection performance. Furthermore, we also consider sensing error performance as a function of the number of SUs. Presented results clearly show that the effectiveness in the detection performance for both JUs and TUs using the proposed deep neural network as compared to conventional cooperative spectrum sensing schemes, such as K-out-of-N, and support vector machine.