With the increase of signal bandwidth demands in these applications, it is a technical challenge to achieve the required high Nyquist sampling rate. In this paper, a new sub-Nyquist sampling array architecture for the multi-band signal carrier frequency and DOA estimation based on a uniform linear array is proposed, which consists of a two-stage estimate algorithm and the corresponding parameter pairing approach. We first adopt a sub-Nyquist sampling structure, namely, a modulated wideband converter (MWC), to sample the received signal at the reference sensor, and then the signal is reconstructed and the carrier frequencies are estimated accordingly. Second, the sub-Nyquist sampling data of other sensors are similarly obtained by using one channel of the MWC, respectively. In doing so, conventional DOA estimation algorithms can be adopted for the multi-band signals and a linearly correlated steering vector can be easily resolved. Furthermore, an effective pairing approach is also developed to match the multi-band signals in the estimated DOAs. The effectiveness of all the proposed methods is verified by simulation results.