Abstract: To solve the problem of blind estimation of spreading sequences for the variable spreading length multi-rate DS/CDMA signals, a new parallel blind estimation algorithm, based on blind source separation, is proposed. Firstly, the method uses the reprocessing of the power spectrum algorithm to estimate the periods users employ different rates of spreading sequences. Secondly, based on period characteristics, the signals are modeled with primary users and interference users, treating users with the same period as primary users and others as interference users. Finally, for synchronous and asynchronous systems, the signals are divided into single or double times of spreading period length, respectively, forming a data matrix. The robust JADE algorithm is used for the blind estimation of the spreading sequences. The computer simulation results show that this method performs well at lower signal-to-noise ratios.