ID 原文 译文
58448 以多用户加权最小均方误差为优化目标，综合考虑天线结构硬件限制、用户公平性和功率限制，采用数字—模拟混合式两级预编码与最优一级数字预编码最大逼近的方法求解两级混合预编码. Aiming at minimizing the multiuser weighted minimum mean square error，an op?timization problem was established under the constraints of antenna structure，user fairness and the limita?tion of power and the hybrid precoding problem was solved by utilizing the best approaching to the digitalprecoding.
58449 仿真结果表明，所提算法能最大限度地减小所求混合预编码矩阵与最优一级预编码矩阵的误差，具有较好的收敛性，可获得近乎最优的混合预编码性能增益. Simulations verify that the proposed algorithm narrows down the gap between the hybrid pre?coder and the optimal one-stage precoder to the greatest extent with a good convergence and obtains anear-optimal solution with significant gain.
58450 为改善零样本图像分类中相似度度量方法的鲁棒性，引入了一种用于零样本分类的度量学习方法. To improve the robustness of similarity metric method in zero-shot learning，a new metriclearning for zero-shot image classification was introduced.
58451 该方法由自编码构成，能在特征对齐后的语义嵌入空间中学习到最优的度量函数，用于计算测试样本特征和类标签的语义特征的相似度; It is composed of autoencoders，which canlearn the optimal metric function in the feature-aligned semantic embedding space.
58452 然后利用近邻思想预测类别标签，进而避免产生不合适距离函数导致的分类错误. The similarity betweentest sample features and the semantic features of the class labels can be calculated by metric function，predicting the label of the class by the neighboring method.
58453 实验结果表明，与传统距离度量的算法相比，所提出的方法降低了识别错误率，在公开数据集 AWA、CUB 和 ImNet-2 上的分类准确率分别达到 94.7% 、63.7% 和 28.59% ;同时表明了语义—视觉的映射方向比相反方向的识别准确率高出2.5% ～ 10.1% . Thus，the classification error caused by inap?propriate distance function is prevented. Compared with the traditional distance metric algorithm，theexperiments show that the proposed method reduces the recognition error rate; the recognition accuracy isimproved to 94. 7% ，63. 7% and 28. 59% on the AWA，CUB and ImNet-2 datasets. At the same time，it was confirmed that the recognition accuracy of the semantic-visual mapping direction was 2. 5% ～10. 1% higher than the opposite direction
58454 为解决导航电文信息速率制约导航系统性能的问题，提出了一种基于码移键控调制和多进制低密度奇偶校验码的新型全球卫星导航系统( GNSS) 信号编码调制技术. To solve the problem of navigation message rate limiting the performance of global navigationsatellite system ( GNSS) ，a new technology of GNSS signal design based on code shift keying modulationand non-binary low density parity check code is discussed.
58455 在 － 130 dBm 的接收电平、误码率不大于 10 － 7 的指标下，该技术可以达到 1 500 bit / s 的有效信息速率，并且具有信号结构简单、与现有 GNSS 信号兼容性强等优势. At a receiving level of － 130 dBm and an er?ror rate of no more than 10 － 7，the information rate can be up to 1 500 bit / s. Meanwhile，it has the ad?vantages of simple signal structure and strong compatibility with existing GNSS signals.
58456 理论分析和仿真结果证明，该技术突破了现有卫星导航信号体制下的信息速率极限，能够克服复杂信号结构造成的难以独立捕获跟踪的问题，在卫星导航系统中有比较重要的实用价值. Theoretical analy?sis and simulation prove that this technology can break through the limit of information rate under the ex?isting GNSS signal system，and overcome the difficulties of acquisition and tracking caused by complexsignal structure. Therefore，it has important practical value in GNSS.
58457 针对实际工程中小子样结构机构可靠性分析的问题，提出了 Bootstrap—支持向量回归( SVＲ) —二阶累计量( SOC) 方法框架和多阶矩虚拟样本容量扩充方法，增强了扩充后样本的客观性; In order to solve the small sample reliability problems for structures and mechanisms in realapplications，an analyzing framework named Bootstrap-support vector regression-second order cumulant( Bootstrap-SVＲ-SOC) method together with a sample size extension method combining multi-order mo?ments that could make the extended samples more objective were proposed.