Wideband spectrum sensing based on Multi-Resolution Bayes classifier for cognitive radio

Finding the frequency locations of the occupied bands is a major challenge in wideband spectrum sensing. The common known method is based on wavelet edge detection. However, the low signal to noise ratio (SNR) may bring many fake edges to the power spectral density (PSD) of the received signal. Furthermore, since the wavelet edge detection works on the assumption that the PSD of the received signal has irregular structures at the edges of the occupied bands, it will fail when the edges are smooth. In response to these problems, a novel method for wideband spectrum sensing based on Multi-Resolution Bayes classifier is proposed in this paper. The proposed method is robustious in low SNR circumstance because Multi-Resolution analysis is used to prevent these fake edges. Based on Bayes classifier, the proposed method can still perform well when the PSD of the received signal is smooth at the edges of the occupied bands. Finally, an approximate threshold selection criterion is developed and the performance of the proposed method based on a Gaussian distribution is discussed. Simulation results are presented to verify the method.