Goodness-of-Fit (GoF) based spectrum sensing of orthogonal frequency-division multiplexing (OFDM) signals is investigated in this paper. To this end, some novel local sensing methods based on Shapiro-Wilk (SW), Shapiro-Francia (SF), and Jarque-Bera (JB) tests are first studied. In essence, a new threshold selection technique is proposed for SF and SW tests. Then, three studied methods are applied to spectrum sensing for the first time and their performance are analyzed. Furthermore, the computational complexity of the above methods is computed and compared to each other. Simulation results demonstrate that the SF detector outperforms other existing GoF-based methods over AWGN channels. Furthermore simulation results demonstrate the superiority of the proposed SF method in additive colored Gaussian noise channels and over fading channel in comparison with the conventional energy detector. Manuscript profile

In this paper, the statistical distribution of the received quadrature amplitude modulation (QAM) signal components is analyzed after propagation in a dispersion uncompensated coherent optical fiber link. Two Gaussian tests, the Anderson-Darling and the Jarque-Bera have been used to measure the distance from the Gaussian distribution. By increasing the launch power, the received signal distribution starts to deviate from Gaussian. This deviation can have significant effects in system performance evaluation. The use of the Johnson s_U distribution is proposed for the performance evaluation of orthogonal frequency division multiplexing in an uncompensated coherent optical system. Here, the Johnson s_U is extended to predict the performance of multi-subcarrier and also single carrier systems with M-QAM signals. In particular, symbol error rate is derived based on the Johnson s_U distribution and performance estimations are verified through accurate Monte-Carlo simulations based on the split-step Fourier method. In addition, a new formulation for the calculation of signal to noise ratio is presented, which is more accurate than those proposed in the literature. In the linear region, the Johnson based estimations are the same as Gaussian; however, in the nonlinear region, Johnson s_U distribution power prediction is more accurate than the one obtained using the Gaussian approximation, which is verified by the numerical results. Manuscript profile