Invariant wideband spectrum sensing under unknown variances

In this paper, we divide a wide frequency range into multiple subbands and in each subband detect whether in a primary user (PU) is active or not. We assume that PU signal at each subband and the additive noise are white zero-mean independent Gaussian random processes with unknown variances. We also assume that at least a minimum given number of subbands is vacant of PU signal and propose an invariant generalized likelihood ratio (GLR) detector. The concept of the grouping of subbands allows faster spectrum sensing of a subset of subbands which may be occupied by a specific PU. Also, we evaluate trade-offs involved in the proposed algorithms by simulation.