Deep Learning based Cooperative Spectrum Sensing with Crowd Sensors using Data Cleansing Algorithm
D. Raghunatha Rao, T. Jayachandra Prasad, M. N. Giri Prasad · 2022 International Conference on Edge Computing and Applications (ICECAA) · 2022
Cooperative sensing is a solution to augment the performance of detection, in which Secondary Users (SU) cooperate with each other for sensing the spectrum to discover the spectrum availability. Spectrum sensing is one of the effective components of Cognitive Radio Networks (CRN). Spectrum sensing allows a CR, which contains information about its spectrum availability as well as the environment. In this paper, the spectrum availability sensing is carried out using the Deep Residual Network test statistic refining scheme (DRN test statistic refining scheme) in which the DRN contains several layers, such as convolutional layer(cony), pooling (pool), activation function, batch normalization, residual blocks and linear classifier. Here, the linear classifier provides the sensing output by considering the fused data and sensing data matrix as an input. In order to improve spectrum availability sensing, data cleansing algorithm along with data fusion have been implemented. In addition, all the sensing processes are carried out in Cooperative Spectrum sensing (CSS) with Crowd sensors. Moreover, the experimental result demonstrates that the devised DRN test statistic refining scheme attained the better detection probability of 0.9578 for the matrix size of 10 X 5 using Rician channel.