A Neural Network Approach to Decision Fusion for Wideband Cooperative Sensing
Ashwini Kumar Varma, Debjani Mitra · 2018
Cooperative spectrum sensing is a widely applied approach in cognitive radio (CR) system to improve the reliability of primary user detection by overcoming the effect of real-time issues such as multipath fading and shadowing. The improvement in the detection performance comes at the cost of cooperation overhead. Here, overhead refers to the huge computational time requirement to perform cooperative sensing which affect the performance of the system. In this paper, an adaptive artificial neural network (ANN) approach has been used for decision fusion that jointly optimizes the time requirement and the performance of the system. The proposed model adapts to the surrounding conditions while training the model, in which higher weights are initialized to the most reliable CRs, while lower weights are initialized to least reliable CRs. The system is validated over real-time scenario using NI-USRP 2942R test bed. Results demonstrate that the proposed method is better than the traditional decision fusion schemes in terms of detection performance.