Blind Spectrum Sensing Method Using T² Statistic and Goodness of Fit

In response to the problem that the performance of the spectrum sensing is poor in the low signal-to-noise ratio and the sampling number, we construct new test statistic. Firstly, the T² statistics is constructed by using the sample feature of the sample test matrix. Then according to likelihood ratio criterion, the new test statistic is contained. Finally, when the spectrum is idle, the probability density function of the test statistic is deduced. And the blind spectrum sensing algorithm based on the Beta distribution is proposed. However, when the number of samples is small, the detection performance is poor. Therefore, according to the Anderson-Darling criterion, a new algorithm based on the Beta distribution and goodness of fit is proposed. The proposed algorithms are simulated under AWGN channel and Rayleigh fading channel, and compared with the simulation results of the existing blind spectrum sensing algorithm. The proposed algorithm has better detection performance and do not need the main user information, no feature decomposition, no influence of noise variance.