Cyclostationary feature based spectrum sensing via low-rank and sparse decomposition in cognitive radio networks
Jincai Du, Hai Huang, Xiao Jing, Xuqi Chen · 2016
Spectrum sensing is a key component in cognitive radio networks, which allows secondary users to communicate without causing harmful interference to primary users. Cyclostationary feature based spectrum sensing has proven preferable to other methods under low signal-to-noise ratio conditions. To detect the presence of primary signals, conventional cyclostationary feature based schemes tend to simply compare the values of signal features to a predefined threshold. However, such schemes would lead to dramatic performance degradation when signal features are overwhelmed by noise. This paper proposes a novel scheme that applies the low-rank and sparse decomposition technique to cyclostationary feature based spectrum sensing. The spectrum correlation function matrix is decomposed into two matrices, of which the low-rank one represents noise and interference while the sparse one represents cyclostationary features of PU signal. Subsequently, the scheme takes advantage of the signal features in the sparse matrix to determine the presence of PU signal. Simulation results have demonstrated the superiority of our proposed scheme in terms of detection probability and false alarm probability.