Leveraging Bayesian Factor Analysis for Cooperative Wideband Spectrum Sensing

This paper introduces a novel cooperative wide-band spectrum sensing (WSS) framework to effectively estimate the occupancy state of multiple frequency bins across a large bandwidth. Notably, this framework operates independently of any prior information about the number and characteristics of primary users (PUs) signals and the sensors' noise power. The key approach is recasting the cooperative WSS problem within a variational Bayes factor analysis (VBFA) framework, which leads to a novel sensing approach capable of detecting unused spectrum bands, estimating sensor noise power even in the presence of PU signals, and counting the number of (PU) transmitting in each frequency band. The framework is applied to a realistic case study examining the effects of path-loss, shadowing, and frequency-selective multipath channels between (PU) and sensors. Numerical results demonstrate that the proposed solution surpasses state-of-the-art algorithms, showing remarkable performance, particularly in low (SNR) conditions-achieving a detection probability of 90% at an SNR of -10 dB. Finally, the effectiveness of the proposed method is validated by comparing its performance with that of a genie-aided likelihood ratio test (LRT)-based detector, revealing that under certain conditions, our solution approaches the performance level of the benchmark.