Blind multiband signal detection with multirate snapshots in low SNRs

Abstract

A novel multirate sampling scheme is proposed for wideband spectrum sensing. The signal is subsampled by shifted parallel channels working at same rate. However, the rate switches among several rates between neighboring snapshots where the spectrum aliases at different frequencies. Basing on this, we propose a blind multiband signal detection algorithm. In this method, we construct frequency locator polynomials (FLPs) to locate the nonzero components of the original spectrum from the subsampled signal for each snapshot. We compute the FLP evaluations, which differs in the snapshots sampled with distinct rates. The evaluations are statistically described by the complex Gaussian distributions for each detected frequency. The detection is then completed by fusing these probabilities contributed by the latest several snapshots. Simulation results show that the proposed algorithm can efficiently lower the false alarms.