In order to estimate the number of coherent sources, a Hankel matrix with the size of half the number of the received arrays is constructed using snapshot data of observed vectors. And the rank of the Hankel matrix is only related with the number of signal sources, no matter the signals are uncorrelated or coherent. We can get the signal and noise eigenvalues by conducting the singular value decomposition (SVD) to the Hankel matrix, the source number can be obtained by calculating the maximum ratio of each eigenvalue pair. The complexity of the algorithm is reduced greatly as only part of the observed data (single snapshot) is used. The Monte-Carlo simulation results demonstrate the feasibility of the algorithm.