Spectrum sensing is a vital task in cognitive radio.In this paper,a new test statistic is constructed via the sample features; the probability density functions(PDF)of the test statistic are derived and a blind spectrum sensing based on Students distribution is proposed.Comparing with the energy detection(ED)and normal goodness of fit(GOF)algorithms,simulations show that the performance of proposed algorithm is much better than the ED with the noise variance as prior information,and overcome the weakness of ED algorithm and GOF algorithm,which are affected by the noise uncertainty and noise variance in the case of Gaussian channel.