Compressive Wideband Spectrum Sensing in Cognitive Radio Systems Using CPSD

Accurate wideband spectrum sensing and primary user detection with high probability in low SNR condition is critical in cognitive radio systems. In these systems, usually, the bandwidth of the spectrum to be sensed is very wide but the spectrum occupancy is very low. Complexity of conventional wideband spectrum sensing methods can be very high, due to a very high sampling rate that is needed. In compressive spectrum sensing a new sampling method is used that can recover sparse signals from under Nyquist sampling rate. Energy detection is the most common way of primary user detection, as its computational cost and implementation complexity are low and no knowledge of primary's signals is needed. However, due to noise variance estimation error, the performance of energy detection method in low SNR condition is poor. In this paper we have used a so called complex valued power spectrum, CPSD, rather than conventional PSD, to overcome this problem. It will be shown that the CPSD of white noise is zero; therefore, the estimation of noise variance is not needed. The proposed method is robust against noise uncertainty and its computational cost is almost the same as conventional energy detection methods. The results have been verified by computer simulation.