Wideband spectrum sensing is a fundamental problem in cognitive radio networks. In this paper, a wideband spectrum sensing procedure which consists of four steps, i.e., coarse support recovery, fine binary detection, band edges detection and clustering, for modulated wideband converter (MWC) based cognitive radio system at sub-Nyquist rate is proposed. Simulations show that with a sub-Nyquist sampling rate, the proposed spectrum sensing procedure can locate spectrum holes correctly.