Model order estimation for sparse wideband signal compressive sensing

Compressive sampling (CS) has been wildly investigated in wideband signal spectrum sensing. A variety of signal reconstruction algorithms are reported for CS. Usually, the eigenvalues of covariance matrix of compressive measurements are used to estimate signal model order. However, due to the noise introduced in the sampling stage and the finite length of sampling data, the performance of signal model order estimation would be degraded. In this paper, we present a model order estimation approach for CS signal reconstruction. The proposed approach is based on minimum description length (MDL) criterion, which is used to distinguish between signal subspace and noise subspace. We evaluate the performance of the proposed algorithm in the framework of modulated wideband converter (MWC), which is designed based on CS. It is shown that the signal model order can be reliably estimated using the proposed algorithm.