Considering the high computational burden of the previous kernel spectrum sensing methods

this paper proposes a computationally more efficient Nystrom subspace matching (NSM) algorithm.Based on the independent identically distributed observations

the subset is randomly chosen to implement the Nystrom approximation and reconstruct the related kernel features in a high-dimensional Euclidean space.Then

the related Nystrom subspaces respectively for the primary users and the secondary users are modified

and the Frobenius range between these two subspaces can be computed to determine whether the primary users exist or not.Compared to the previous kernel subspace matching methods

the novel version reduces the computational complexity by 66% while provides almost the same detection performance.Computer simulations are conducted to evaluate the performance of the proposed algorithm.