Blind sequences estimation methods, which work without prior knowledge about the transmitter, have been proposed in [1]-[3] ; but they are restricted to a single-user context. Here, we extend the approach originally proposed in [2] to the multi-user context. We have developed a fast and efficient blind synchronization criterion based on the behavior of the FROBENIUS square norm of a submatrix that scans the intercepted signal correlation matrix. By using eigenanalysis techniques and theoretical linear algebra results, we prove that the maxima of this criterion correspond to the synchronization times with transmitters. Then,eigenanalysis techniques allow to estimate the spreading sequences. Eigenanalysis was associated with additional processing modules aimed at solving dephasing problems due to the transmission channel.