：为解决无线分集相干光接收机的自适应盲检测问题，提出了一种新的离散时间连续状态的网络输出反馈偏置型的复Hopfield神经网络用以解决多值QAM信号的盲检测问题。反馈电压偏置的引入即不脱离传统Hopfield模型，又能有效满足多值信号检测时所需的搜索空间变大的特殊要求。全文完成多值信号盲检测的优化问题构造和能量函数的映射，给出能量函数的证明、分析和它的约束条件，给出适用该问题的激活函数的基本特征，正确盲检测信号的权矩阵的配置方法。最后，通过详细的仿真结果展示和与其他算法性能对比进一步验证算法的有效性和优越性并指出算法所存在的问题和下一步的研究方向。

To solve the special issue of electrical adaptive blind equalization in wireless spatial diversity optical coherent receivers, a new blind detection algorithm of multi-value QAM signals using output-feedback-bias(OFB) type complex discrete-time continuous state(DTCS) Hopfield neural network was presented. The OFB will not change the traditional Hopfield model. The proposed OFB-DTCS Hopfield neural network can meet special requirement of the multi-valued signal detection which need enlarger the search space. The blind detection problem of multi-valued QAM signals was transformed into solving a quadratic optimization problem. How to map the cost function of this optimization problem to the energy function of OFB-DTCS Hopfield neural network was also shown. The proof, analysis and its constraints of the energy function were shown, respectively. A complex activation function to fit this special problem was discussed. Then a special connective matrix was constructed to ensure algorithm detect signals correctly. Finally, detailed simulation results and performance comparison with other algorithm were shown to demonstrate farther the effectiveness, superiority and shortage of this new algorithm.