In view of the difficulties in the blind estimation of the Pseudo-random Noise（PN） sequence OVSF code for WCDMA uplink signals in low Signal to Noise Ratio（SNR）

an approach is proposed by using the correlation matrix and cumulative average of the signals combined with its singular value decomposition.When such parameters of the WCDMA signals as the OVSF code period and rate are known

the received signals are periodically divided into multiple continuous observation vectors based on the OVSF sequence period with a random fixed value as the starting point for the correlation matrix and cumulative average.Then the OVSF sequence of the WCDMA signals is blindly estimated from the main component eigenvector and the signals are despreaded.Simulation results show this method works well in a lower SNR scenariors.