Aiming at the problem of exorbitant sampling rate and sensing time too long in wideband sensing, a wideband spectrum sensing method using improved multiple signal classification (MUSIC) based on modulated wideband converter (MWC) was proposed. The MWC for sub-Nyquist sampling was used, the number of signals by minimum description length (MDL) criterion was estimated, and the signals’ location using improved MUSIC was estimated. In the improved MUSIC, an adjustment factor was given, which enhanced the MUSIC spectrum at the signal’ locations and reduced others, and the noise jam was restrained. An amount of calculations was reduced, because it neither needed to recover original wave, nor calculated PSDs in the whole process. The complexity of sensing method was small, so that it increased sensing efficiency. The results show a reliable detection even in low signal noise ratio.