Cooperative cognitive spectrum sensing based on optimized time-frequency signal analysis

Spectrum sensing is used in cognitive radio to detect the free portions of spectrum in a licensed frequency band. We introduce a cooperative spectrum sensing scenario in which the decisions from the secondary users are combined for better sensing accuracy. Each secondary user sends its decision to a central node which combines all individual decisions. A discrete Fourier transform (DFT) lter bank based architecture is used by each seconda user for ef cient detection of a primary user signal in a desired time-frequency slot. The prototype lters underlying the DFT lter banks are optimized to provide maximum timefreque resolution. We formulate an objective function to represent the time-frequency distribution of signal energy and use numerical methods to obtain optimized prototype lter To address the problem of noise power uncertainty in cognitive radio systems, we introduce a method for denoising the received signal which is based on goodness-of- t statistical te. We compare the performance of the proposed method with other spectrum sensing methods in terms of receiver operating characteristics (ROC). The spectrum sensing performance is also analyzed in the presence of noise power uncertainty. Finally, the hardware implementation aspects of the proposed architecture are also analyzed using a eld programmab gate array (FPGA).