基于神经网络的低信噪比CBOC信号组合码序列盲估计

Neural network approach to blind-estimation of combined code sequence in lower SNR CBOC signals

针对低信噪比下复合二进制偏移载波(composite binary offset carrier,CBOC)信号的组合码序列盲估计问题。首先采用奇异值分解(singular value decomposition,SVD)的算法对CBOC的组合码序列进行可行性验证,可得在已知相关参数的情况下对CBOC信号组合码序列盲估计是可行的;其次就SVD在长序列估计中计算量和存储量需求大的问题,进一步提出主分量神经网络解决上述问题,同时引入最优变步长收敛模型改善神经网络(neural network,NN)收敛速度。利用无监督NN的自适应主分量提取信号特性,避免批处理运算,实现CBOC信号组合码序列盲估计。实验表明,NN能在-20dB下达到精确估计序列的目的,且算法有稳定性高、复杂度低、收敛速度快等优点。

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.