Compressive Wideband Spectrum Sensing in Cognitive Radio Systems Based on Cyclostationary Feature Detection

Abstract

High precision spectrum sensing is a critical component in cognitive radio systems. This is more critical when our interested bandwidth is very wide in noisy channel environments. There are many detection ways for spectrum sensing, but each of them has their problems. In this paper we use cyclostationary feature detection which is robust against noise uncertainty, but it needs very high sampling rate, especially when the interested frequency band is wideband. Hence its computational and hardware cost are high, Compressive sensing is a new sub-Nyquist sampling method, which asserts can completely recover specific signals, which are sparse in a certain domain. This paper helps to reduce the required sampling rate of cyclic detector by using the compressive sensing procedure and exploiting the sparsity of the cyclic features in the two-dimensional cyclic spectrum domain. In addition this paper proposes new scheme for reformulating the linear relationship between the compressive samples acquired in frequency domain and the two-dimensional cyclic spectrum. Simulations show that the proposed spectrum sensing scheme can reduce the required sampling rate with little performance loss, and is robust against noise uncertainty in low SNR conditions, also show that the reconstruction accuracy and probability of detection for proposed scheme is higher than for existence methods.