Cyclostationarity-based wideband spectrum sensing using random sampling

Abstract

Cognitive radio (CR) systems offer higher spectrum utilization by opportunistically allocating the unused spectrum from primary users to secondary users. For CR it is vital to perform fast and accurate spectrum sensing in a wideband and noisy channel. Cyclic feature detection performs well in signal detection and is also highly robust to noise uncertainty. However, it requires a high sampling rate when operating over a wideband channel. Based on the sparsity of the cyclic spectrum, compressive sampling technique can extend sparse reconstruction to its case. This paper develops a simpler cyclic spectrum recovery method based on random sampling and demonstrates faster and better performance. Recent research on discrete random sampling provides a new connection between sub-Nyquist sampling and aliasing as a noise floor that can be dynamically shaped by different distributions of sampling times. Practical analog-to-digital converters can implement these random sampling schemes. Thus, a reduced hardware complexity cyclic feature detector based on the reconstructed cyclic spectrum is proposed to identify the spectrum occupancy within the entire wideband.