No abstract

Multiband spectrum access plays an essential role in cognitive radio systems so as to increase the network's throughput through wideband spectrum sensing. It includes identifying the number of subbands comprising a wide spectrum by edge detection, and also examining their occupancy through primary user detection techniques. Despite the offered accuracy of the wavelet-based approaches, their complexity becomes a drawback. Remarkably, the features revealing property of cepstral analysis and its implementation simplicity make it a suitable candidate for signal detection. Motivated by these reasons, this paper presents a wideband spectrum sensing approach based on cepstral analysis. First, we propose the differential log spectral density algorithm for the edge detection phase in order to detect the spectral boundaries within the wideband of interest. Also, we present a mathematical framework of the proposed algorithm and an expression for the detection threshold of the proposed detector is derived. The simulation results have showed a superior performance of the edge detection algorithm to different wavelet-based techniques at low-to-medium noise power. Used in conjunction with denoising, the proposed edge detector shows good detection results at low signal-to-noise ratio. For the primary user detection phase, we introduce the improved passband autocepstrum detector to tackle the misdetection problem of noise-like signals and it outperforms different state-of-the-art techniques. Finally, the uncertainty problem of the subbands center frequencies is addressed and the baseband autocepstrum detector is introduced as a potential solution to improve signal detection in frequency selective fading.

In this manuscript, we introduce a semi-blind spectrum sensing technique based on cepstral analysis for interweave cognitive systems. The misdetection problem of spread spectrum signals leads to erroneous sensing results, which affect the quality-of-service of a legitimate user. The simplicity and accuracy of cepstral analysis approaches make them reliable for signals detection. Therefore, we formulate the averaged autocepstrum detection technique that utilizes the strength of the autocepstral features of spread spectrum signals. The proposed technique is compared with the energy detection and eigenvalue-based detection techniques and shows reliability and efficacy in terms of detection accuracy.