Spectrum Sensing Based on Stochastic Resonance and SOM Neural Network

To address the issue of decreased detection performance in low signal-to-noise ratio (SNR) environments and enhance the accuracy of spectrum decisions, a spectrum sensing method based on monostable stochastic resonance and self-organizing map (SOM) neural network (MSRSOM) is proposed in this paper. Firstly, we introduce the monostable stochastic resonance to enhance the SNR of sensing signal, and improve the perception ability of weak signal. Next, based on information geometry (IG) theory, the signal features are extracted by the kullback leibler (KL) divergence, which shows the information difference between signals on the manifold directly. Finally, the signal features are classified and spectrum sensing is achieved using the SOM neural network. To evaluate the performance of MSRSOM, it is compared with other widely used spectrum sensing algorithms. The simulation results demonstrate that MSRSOM significantly enhances the detection performance of spectrum sensing systems in low SNR conditions.