FPGA implementation and performance study of spectrum sensing based on entropy estimation using cyclic features

This work presents a spectrum sensing technique based on the entropy of frequency domain autocorrelation of receiving signal at different cyclic frequencies. The performance of the proposed sensing technique is compared with other sensing techniques such as energy detection using Bayesian and Neyman–Pearson criteria, entropy estimation under frequency domain, cyclostationary feature detection. The performance of sensing algorithms is also analyzed for single node and multinode/cooperative environment under most probable channel effects such as fading, shadowing, receiver’s uncertainty and free space path loss using Monte-Carlo methods. Simulation results reveal that the proposed sensing technique is able to detect signals of signal-to-noise ratio up to −24 dB with five nodes in cooperation while maintaining a false alarm probability of 0.1 and a detection probability of 0.9. The proposed sensing algorithm is also implemented in Virtex-4 Field Programmable Gate Arrays.

Figure optionsDownload as PowerPoint slideHighlights
► A Spectrum sensing method based on entropy estimation using cyclostationary features.
► It outperforms energy, cyclostationary and entropy detection techniques.
► Proposed method detects signals up to −19 dB SNR with single node.
► It detects signals up to SNR −24 dB with five nodes in cooperation.
► The algorithm is implemented in Xilinx Virtex-4 Field Programmable Gate Array.