In this paper, a modified MUSIC algorithm, namely, VA-MUSIC is presented for the direction-of-arrival (DOA) estimation of noncircular signal sources. By using the conjugate counterpart of the array output, a virtual array is formed to extend the equivalent array aperture, so that it can handle more sources than sensors. Then, based on dividing the virtual array into many subarrays, and composing a new covariance matrix, the new algorithm is capable of estimating both uncorrelated and correlated signals simultaneously. Moreover, the performance of the new DOA algorithm is investigated and compared with the Cramer-Rao bound (CRB). Finally, computer simulation results are given to demonstrate the effectiveness and superiority of the proposed algorithm.