Spectrum sensing in satellite cognitive radios: Blind signal detection technique

In this article, cyclostationary analysis of the communication satellite signals has been performed and a simpler spectrum sensing technique using the features in the spectral density has been devised. The proposed technique makes use of the time smoothing cyclic periodogram for estimating the spectral density of the captured signal and a normalized feature vector is calculated from the spectral density. Using the defined threshold, feature vector peaks are counted and decision of signal presence or absence is made. The proposed technique is simpler in operation as compared to other techniques, which requires high resolution analysis for feature detection, particularly pattern recognition using artificial neural networks training datasets or wavelet analysis for de-noising spectral density. The proposed technique works without any prior knowledge of the captured data. It is also found to be least affected by the location and magnitudes of the feature peaks as they appear in the spectral density due to maximum value normalization. Detection of the signals has been done without any pre-processing such as de-noising. The USRP2 platform is used to capture the real-time satellite modulated carriers of BPSK, QPSK, and 16-QAM along with the noise in the satellite spectrum using GNU Radio. Further analysis and simulations are performed in the Matlab. The next generation of satellite cognitive radio designs can use this new algorithm to quickly sense, at low resolution and with less complexity/stability in the presence of a carrier signal or noise in the satellite spectrum. (c) 2016 Wiley Periodicals, Inc. Microwave Opt Technol Lett 58:1377-1384, 2016