Wavelet based Non-Reconstructive Compressed Spectrum Sensing against Non-Gaussian Noise

Abstract

In cognitive radio, spectrum sensing in non Gaussian channel is a challenging task. In addition to non Gaussian noise channel, wideband sensing demands computational complex signal processing algorithm for spectrum sensing. Compressive sensing (CS) has emerged as one of the promising signal processing techniques for spectrum sensing. In CS-based sensing, the sensing is performed on the reconstructed signal that adds extra computational overhead. To mitigate the computational overhead, we present a non-reconstruction-based spectrum sensing technique that uses the energy of the discrete wavelet transform (DWT) coefficients of the compressed measurements in non Gaussian channel, as test statistics. We compare the performance of the proposed algorithm with the conventional energy detection against both non-Gaussian and Gaussian noise. Numerical simulation is carried out using a 5G Universal-Filtered Multi-Carrier (UFMC) signal. The simulation results demonstrate that the DWT-based sensing achieves nearly −8 dB and − 8.5 dB SNR-wall against non-Gaussian and Gaussian noise, respectively. In contrast, the compressed DWT achieves −4.5 dB SNR-wall at a much lower computational complexity than conventional techniques.