加密系统中Alice端通过光纤传送相干态量子信号给Bob端,得到的一系列相关的高斯连续变量,然后从中提取出二进制密钥,这一过程称作高斯连续变量量子密钥分发。其中的数据协调步骤,即对相关连续变量的错误校正,是本文的研究问题。本文在分层错误校正协议(SEC)和多电平编码／多级解码(MLC／MSD)协议的基础上,用基于LDPC码的Slepian-Wolf编译码器实现协调。根据连续变量的多级Tanner信息传输流图,本文推导了多级迭代的内信息和外信息置信传播(LLR-BP)更新公式,并用实验比较了MSD三种译码模式的收敛性质及性能。仿真结果表明该算法可在信道信噪比2.5 dB以上实现10000个连续变量序列的可靠协调,协调效率可达92.3％。

The party Alice in a cryptograph system transmits coherent quantum signals to the party Bob, obtaining a series of correlated continuous Gaussian variables. The process to distill a binary key out of the correlated Gaussian variables that they share is called continuous-variable quantum key distribution (CVQKD). Reconciliation in CVQKD, i.e., to correct differences between the correlated variables, is the key issue discussed in the paper. Based on both sliced error correction (SEC) protocol and multilevel coding／multistage decoding (MLC／MSD), reconciliation is implemented by LDPC code-based Slepian-Wolf CODEC. According to the proposed multilevel Tanner graph which depicts message transition of information, multistage iterative intrinsic and extrinsic information belief-propagation (LLR-BP) update formulae are derived. Three decoding modes of MSD are compared experimentally in convergence and performance. Simulation results show that the proposed algorithm can reconcile reliably 10000 continuous quantum variables with efficiency of 92.3％ when SNR of the channel is above 2.5 dB.