Focusing on the problem of blindly estimating the combined code sequence of the composite binary offset carrier (CBOC) signal under low signal to noise ratio, this paper first adopts the algorithm based on singular value decomposition (SVD) to verify the feasibility of the CBOC combined code sequence, obtaining the result that given relevant parameters it is feasible to estimate blindly the combined code sequence of the CBOC signal. Second, focusing on the problem that the SVD algorithm needs too much calculation and storage when estimating long sequence, this paper proposes principalcomponent neural network (NN) as the solution, and meanwhile introduces the optimal variablestep convergence model to improve the convergence rate of NN. Using the selfadaptive principalcomponent of the unsupervised NN to extract signal peculiarity, and avoiding processing batch, can thus realise the blind estimation of the combined code sequence of CBOC signals. Simulation experiment indicates that the NN algorithm can estimate sequence accurately under an SNR at -20 dB, and holding advantages like high stability, low complexity and high convergence rate.