Implementation issues in cooperative spectrum sensing with soft fusion

Cooperative strategies are known to be effective spectrum sensing techniques capable to exploit spatial diversity to increase the detection performance and counteract the hidden node problem. The main cost related to cooperation is given by the sensing data that must be exchanged among the cognitive radio (CR) nodes and the fusion center (FC). In this paper we focus on soft fusion (SF) techniques considering in particular two approaches: eigenvalue based and energy detector (ED) based cooperative sensing. These strategies are analysed considering all the relevant impairments that can be encountered in a realistic scenario: path-loss, shadowing, frequency-selective fading, and finite data representation between the CR nodes and the FC. Our study reveals that in scenarios of practical interest, the sensing data can be quantised and represented with 4-6 bits per sample without appreciable performance degradation. We also demonstrate that although eigenvalue based detectors can exhibit superior performance in presence of a far field spectrum user, in scenarios in which the path loss differences among the CR nodes are significant, ED based techniques must be preferred.