Compressive Joint Angular and Frequency Spectrum Sensing Based on MUSIC Spectrum Reconstruction

Abstract

This paper aims to introduce a new tool for wideband spectrum sensing in cognitive radio applications based on the multiple signal classification (MUSIC) two-dimensional (2D)-pseudospectrum, which is able to locate the active sources [licenced users (LUs)] or other secondary users in both the angular and frequency domains. The proposed tool consists of two-stages compression and reconstruction procedures. The spatial-domain compression is introduced by activating only some of the antennas in the underlying uniform linear array (ULA) leading to a non-ULA of active antennas. The time-domain compression is introduced by allowing the receivers connected to every active antenna to sample the analog signals below the Nyquist rate. This compression aims to alleviate the burden on the analog-to-digital converter especially when a broad frequency band has to be sensed. The simulation study shows that the constructed MUSIC 2D-pseudospectrum can be used to locate the direction-of-arrival (DoA) and the frequency band of the LUs with a much better resolution than the one that can be afforded by the already proposed 2D power spectrum. Note that the number of sources is beyond that of active antennas and the sampling rate in each receiver is only 17 / 42 times the Nyquist rate. Moreover, the evaluation on the root mean square error in terms of DoA estimation indicates that the proposed MUSIC 2D-pseudospectrum construction approach outperforms the already proposed 2D power spectrum reconstruction method. In fact, the performance of the proposed approach based on the sub-Nyquist-rate samples is only slightly below the approach that is based on the pseudospectrum constructed from Nyquist-rate samples. Furthermore, the evaluation on the detection performance of the spectrum sensing tool based on the MUSIC 2D-pseudospectrum underlines the promising benefits of the proposed approach.