Cognitive radio (CR) is one of the forthcoming technologies for efficient spectrum utilization. It is critically dependent on fast and accurate spectrum sensing which becomes challenging for wideband systems that typically suffer from low and variable signal to noise ratio (SNR) conditions. For wideband sensing, simple energy detection is usually augmented by the double threshold method using forward consecutive mean excision (DT-FCME) algorithm, or approached by eigenvalue detection (EVD). For both these methods, it is difficult to get both high accuracy and fast sensing at same time. In this paper, an algorithm named as correlation of inverted spectrum segments (CISS) is proposed which computes in a specific manner, the crosscorrelation of intra sets of successive sub-bands in the spectrum of wideband sensing data. The approach is significantly faster than both EVD and DT-FCME and maintains high detection probability. In addition, the detection performance under low and variable SNR is also better than the conventional methods used in wideband sensing. The algorithm is analyzed mathematically, and applied to synthetic and real time GSM data captured from an universal software radio peripheral (USRP) setup. The algorithm would be useful as a robust and fast sensing scheme for wideband CR system.

Keywords: Cognitive radio, Wideband spectrum sensing, Variable noise floor, NI-USRP 2942R