ID 原文 译文
57748 仿真结果表明，与传统的协方差差分算法、噪声协方差矩阵估计算法、秩迹最小化算法以及稀疏重构算法相比，所提算法不仅能较好地抑制非平稳噪声的影响，而且在低信噪比、低快拍数情况下具有较强的稳健性和较高的估计精度. Simulation shows that，compared with the traditional covariance difference algorithm，the noise covariance matrix estimation algorithm，the rank trace minimization algo- rithm，the sparse reconstruction algorithm，the proposed algorithm can not only suppress the influence of non-stationary noise effectively，but also has strong robustness and high estimation accuracy under low signal noise ratio and low snapshot number.
57749 为了优化空间耦合低密度奇偶校验( SC-LDPC) 码在加性高斯白噪声信道下的性能，改变了其边缘扩展规则，将规则 SC-LDPC 码原模图上变量节点的边连接到与其相对应的前一个位置的校验节点上，提出了一种新型的空间耦合结构. In order to optimize the performance of spatially coupled low density parity check ( SC-LDPC)codes over the additive white gaussian noise channel，a novel spatial coupling structure was proposed bychanging the edge spreading rules，the edges of the variable nodes were connected to the check nodes ofthe previous position.
57750 对其基矩阵、码率、度分布及译码复杂度进行分析，结果表明，与传统的 SC-LDPC 码相比，链长相同时，新结构的码率有所提高，校验节点的度分布变大. Basis matrix，rate，degree distribution and computational complexity were analyzedand the results show that the rate of the novel structure is promoted with the degree distribution of checknodes become larger.
57751 最后通过 BP 译码验证了提出的空间耦合新结构译码性能的优越性. Finally，the performance of the proposed novel structure is verified by BP decoding.
57752 针对无源目标分类系统中精度和费用之间不平衡、采用手工提取特征的方法进行特征提取工作量较大的问题，提出了一种基于误差逆传播( BP) 神经网络的信道状态信息( CSI) 无源目标分类方法. Aim at the imbalance between accuracy and expense，the heavy workload of manually extrac- ting features in current device-free target classification systems，a channel state information ( CSI) de- vice-free target classification method based on error back propagation ( BP) neural network is proposed.
57753 通过提取 WiFi 信号的CSI 作基信号，并结合具有自主学习数据特征能力的神经网络方法，设计了 BP 神经网络的训练模型，减少了手工提取特征带来的开销. By extracting the CSI of the WiFi signal as the base signal and combining the neural network method with the characteristic of autonomous learning data features，the BP neural network training model is designed， which reduces the overhead caused by the manual extraction feature.
57754 实验结果表明，以身高分类为例，所提方法能够区分 4 个不同身高段，且平均分类准确度可以达到 90% 以上. Taking the height classification as an example，an experiment is carried out，and it is shown that the proposed method can distinguish four different height segments，and the average classification accuracy can reach more than 90% .
57755 已有边缘缓存技术研究假设已知固定的全局流行度，忽略了反映基站接收到的内容请求历史中的流行度地域差异特性和动态特性，为此，提出了一种基于本地内容流行度预测的内容部署策略. Considering the problem that most works on content placement so far consider global populari- ty，neglecting the demand difference between base stations ( BSs) ，a content placement scheme based on similarity between small base stations ( SBSs) and local content popularity prediction considering popular- ities’geographical diversity are proposed.
57756 首先，考虑流行度的地域特性，将内容请求历史记录相似的小基站分簇; Firstly，SBSs that possess similar historical content requests is identified by similarity measurements.
57757 然后，使用线性回归方法预测每个小基站簇群的本地内容流行度，基于预测的本地内容流行度，利用随机几何和凸优化理论求得次优内容部署决策; Then the probabilities of future requests are predicted for each sim- ilar SBS group utilizing linear regression method. Based on this local popularity，the sub-optimal content placement decision is made according to stochastic geometry and convex optimization.