Abstract: Spectrum sensing technology had been widely concerned in order to find the available spectrum holes to improve the spectrum utilization. Because of the existing algorithms that based on the eigenvector involved a large number of eigenvalue decomposition, so they didn't fit for the real-time detection. Utilized the orthogonality between the signal subspace and the noise subspace, the proposed spectrum sensing algorithm projected the received signal of the secondary users(SU) on the two subspace and achieved fast spectrum perception based on the differences of the projection values. The theoretical analysis and the simulation results show that the proposed algorithm reduces the computational complexity effectively and it is robust to the noise uncertainty and need neither the prior knowledge of the primary user(PU) nor the noise variance. Furthermore, it can achieve a better detection performance with low SNR and small sampling.