Employing sample features for blind spectrum sensing algorithm

In order to overcome the influence of noise uncertainty and noise variance, a new test statistic is constructed by the use of the sample features. In this paper, the probability density functions (PDF) of the test statistic under free of frequency channel and busyness of frequency channel are derived and a blind spectrum sensing based on F distribution is proposed. Unfortunately, its threshold is affected by the number of the samples. Hence a goodness of fit (GOF) based on the F-distribution (GOF-F) is proposed via Anderson-Darling criterion. Finally, with comparison to the energy detection and normal GOF algorithms, simulations show the performances of proposed algorithms are much better than energy detection that the noise variance had been known, and to overcome the weakness of energy detection and GOF algorithm, which were affected by noise uncertainty and noise variance in the case of Gaussian channel.