Cyclostationary feature-based compressive sensing in cognitive radio networks
Spectrum sensing is an important function of the cognitive radio (CR) system and is designed to detect the primary users (PUs) signals. The high accurate cyclostationary feature detector requires the high sampling rate and imposes heavy computational loads to the system. Then Compressive Sensing (CS) is applied in the CR spectrum sensing for its low sub-Nyquist sampling rate. However typical CS methods still have a high computation overload due to their complicated recovery algorithms. In this work, we exploit CS in a specific sparse domain (i.e., cyclic domain) and propose a cyclostationary feature based spectrum CS scheme to perform the spectrum analysis. The measurement matrix is generated from both the cyclic feature and sparsity prior knowledge. Therefore, a dynamically compressive signal processing (CSP) detector without the signal reconstruction is developed. By applying this novel spectrum sensing detector, we achieve high detection accuracy and robustness to the noise uncertainty as well as reduced energy consumption (due to the low sampling rate). Moreover, the low complexity and short sensing delay are achieved from this non-reconstruction method. After testing in a several systems, the experimental results show that the proposed cyclic CSP-based sensing method is robust to the noise and the new detector outperforms other traditional detectors in the real applications.