A robust two-stage spectrum sensing method using filter bank based energy detector and fourth order cumulants for cognitive radios

Recent years have shown a growing interest in the concept of cognitive radios (CR), in order to access the electromagnetic spectrum in an opportunistic way. Such systems require efficient spectrum detectors able to work in low signal-to-noise ratio (SNR) environments, with little or no information about the signals. In this paper, using the theory of higher-order statistics (HOS), a two-stage spectrum sensing method using filter bank based energy detector and cumulants, is proposed. The method first uses a filter bank, which separates the spectral region of interest into narrow frequency bands (channels). Each subband of the filter bank is then followed by an energy detection block. The normalized power values computed in each time slot of sensing for all the subbands are used as weights for the fourth order cumulants estimated in those subbands. By employing such weights, the proposed sensing scheme is shown to perform better than the conventional cumulants method. Moreover, no a priori information about the waveform of the signal is needed. Simulation result shows 40% improvement in the detection probability of signals at a low SNR of -20 dB for a constant false alarm rate of 0.1 when compared with the conventional cumulants based method.