In this paper, an approach of eigen-structure analysis for DS-SS (direct sequence spread spectrum) signals under NBI (narrow band interference) is proposed, which can estimate the PN (pseudo-noise) sequence blindly in lower SNR (signal-to-noise ratio) and SIR (signal-to-interference ratio) DS-SS signals. Of course, some parameters of DS-SS signals (such as period and chip interval of PN sequence) need to be known. First, the received signal is divided into continued nonoverlapping temporal vectors according to the period of PN sequence, and then we calculate and accumulate the correlation matrix of these vectors one by one. An operation of eigenvalue decomposition may be applied to the matrices; we can estimate the eigen-structure (which include the NBI eigen-waveforms and PN sequences) of received signals from the principal component eigenvectors blindly in the end. Based on the estimated NBI eigen-waveforms and PN sequences, the NBI can be rejected, and the DS-SS signals can be de-spreaded without the PN sequence too. Theoretic analysis and experimental results show that the approach is very effective. It can work well on the lower SNR and SIR ambient.