A Multi-cycle Spectrum Sensing for OFDM signals under Cyclic Frequency Offsets in Cognitive Vehicular Networks

Spectrum sensing plays a key role in Cognitive Radio (CR) technology to acquire information on the occupancy status of the channel. Cyclostationary spectrum sensing is considered one of the most promising approaches. However, its performance is severely degraded by a mismatch between the actual and the nominal cyclic frequency (i.e., the cyclic frequency offset - CFO). This is particularly true for vehicular networks, due to the high mobility and particularly to the presence of a non zero radial acceler- ation between the transmitter and the receiver, that makes the signal filtered out by the Doppler channel chirped, thus introducing the CFO. Consequently, the signal to be detected has to be properly modelled as a Generalized almost-cyclostationary (GACS) process. The proposed approach performs a maximum likelihood estimation (MLE) of the CFO jointly to the detection exploiting multiple cycles, through a maximization of the Generalized Likelihood Ratio Test. The closed form of false alarm probability is derived highlighting that this approach represents a Constant False Alarm Rate detector. Numerical evaluations are provided to show the correctness of the theoretical analysis and that the performance is approximatively constant for a large range of CFO values. Moreover, comparisons with existing algorithms are provided.