In direct sequence spread spectrum systems (DSSS), two types of short or long code pseudo noise (PN) sequence are used to spread the signal. In receivers that extract signal content blindly, the spreading factor is unknown. In this article, the problem of estimating the spreading factor in DSSS signals with frequency offset is discussed. At first, the data correlation matrix is calculated based on the overlapping time intervals method. Then, using correlation matrix, eigenvalues and eigenvectors are calculated. Finally, using the second power of the eigenvector power spectrum density, the value of the frequency offset is estimated. In the next step, using the missing data method, the covariance matrix of these data is constructed for the length of different time windows at different delays. Then, by removing the main digonal of the covariance matrices, the effect of noise is reduced. In the next step, the second power of the Ferbenius norm of the covariance matrices is calculated and the variance value of these norms is the criterion for detecting the spreading factor. The simulation results show that the proposed method presented in this article is highly efficient at low SNRs and has better results than the existing estimators.