Subspace-Based Cooperative Spectrum Sensing for Cognitive Radios

Accurate and efficient spectrum sensing is critical in providing the cognitive radios with radio environmental awareness in order to improve the efficiency of spectrum utilization. Since narrowband sensing techniques only concentrate on one band at a time with prefixed band locations and bandwidth, wideband sensing is relatively more efficient. Considering the adverse effects of shadowing and multipath fading, in this paper, a cooperative wideband spectrum sensing algorithm based on the subspace method is proposed to detect the presence of a number of primary users in the band of interest. Specifically based on the collected samples of the received signals over multiple antennas, each secondary user estimates the number of primary user signals and their carrier frequencies using the subspace method. Before fusing all local estimates, the fusion center needs to determine which estimates belong to which primary users. The k -means algorithm built on the minimum description length principle is proposed for the data association problem, which can further eliminate false alarms. A linear unbiased estimator is proposed for data fusion and it reduces to a weighted sum of local estimates that belong to the same primary user. Experiments are conducted to demonstrate the efficiency of the proposed algorithm in detecting the correct number of primary users and estimating their carrier frequencies.