Novel spectrum edge detection techniques in wideband spectrum sensing of cognitive radio

In this paper, we first propose to apply the dB-scale values of power spectral density (PSD) instead of linear scale values in current edge detection methods. We show that this modification results in almost the same estimation variance in different subbands with various levels of energy. The simulation results show a significant improvement of over 50% in detection rate by applying the modification under certain conditions. Then a new method based on normalization of edge vector is introduced to mitigate the fluctuation effects in edge detection. Finally, to smooth the fluctuations of the estimated PSD and detect the noise level, another new edge detection algorithm based on the window averaging of the estimated PSD is proposed. Simulation results show better performance for the proposed methods compared to the current edge detection techniques.