A new method based on average correlation and correlation decision function(CDF) is proposed to solve the problem of large computational complexity, for estimating the spreading sequence of direct sequence spread spectrum(DSSS) in this paper. Firstly, the signal is divided into non-overlapping data segments by using the sliding window function with the window length equal tothe pseudo noise(PN) code period. The data matrix is arranged in the order of each segment data as one row vector. Then, the maximum value of the average inner product of the corresponding matrices, which is obtained according to the different starting point conditions, can be used to determine the PN code synchronization starting point. On the basis of the realization of PN code synchronization, the symbolic relation between each symbol of the PN sequence and the first symbol is estimated by using the correlative decision function, so as to realize the blind estimation of the pseudo sequence. The effectiveness of the proposed method is showed by theoretical derivations. The performance and the computational complexity are compared between the proposed method and the common methods based on the eigenvalue decomposition(EVD) and triple correlation function(TCF) through simulation comparison experiments. It turns out that, when SNR\geqslant3$, using the CDF method proposed in this paper can get an accurate estimate, and the computational complexity is greatly reduced compared with methods based on the EVD and TCF.