ID 原文 译文
2903 该结构中每层卷积层提取到的特征与后续所有卷积层的输入相连接，增强了网络的特征信息流动性和梯度的反向传播效率，使得模型提取到的行人特征更具判别力。 In this architecture, features from each convolution layer are con-catenated with all subsequent convolution layers so that it can improve the flowability of feature information and the efficien-cy of backpropagation gradient. In this way, the proposed network can learn more discriminative features.
2904 采用多损失函数组合方式训练网络模型，更好的区分行人类别。 Moreover, a combi-nation of three loss functions is used to train proposed network.
2905 最后利用欧氏距离对行人特征相似性进行排序。 Finally, the identification accuracy is obtained by ranking thesimilarity of extracted features measured by Euclidean distance.
2906 在标准行人再识别数据集 VIPeR 和 CUHK01 上的实验表明，本文方法进一步提高了行人再识别精度，并有效改善了深度神经网络的过拟合现象。 Experimental results on the VIPeR and CUHK01 datasetsdemonstrate that MiF-CNN outperform most of existing methods of person re-identification and reduce overfitting more ef-fectively.
2907 传统基于接收信号强度的物理层密钥生成方案在窃听者靠近合法方时，合法方的密钥易被窃听者获取。 The existing physical layer secret key generation scheme is not sufficiently secure when the eavesdropper keeps close to the legitimate user to get the correlated channel characteristics.
2908 针对该问题，在分析密钥误比特率的基础上，提出一种基于信号传播特性的物理层密钥生成方案。 To solve the problem, this paper analyzes thesecret key bit error rate and proposes a physical layer secret key generation scheme based on signal propagation characteris-tics.
2909 方案根据接收信号强度的实测样本估计大尺度衰落模型，提取出多径效应影响下的小尺度参数量化生成密钥。 We calculate the large scale fading model which parameters is fitted by the measured sample, in order to get the smallscale parameter of received signal strength indication under the effect of multipath fading to quantify into binary secret bit.
2910 实验结果表明相比于传统方案，本方案在室内环境窃听距离大于 0.6 倍波长以后，窃听方密钥误比特率大于 0.48; The experimental results show that compared with the traditional scheme, in the indoor environment, after eavesdropping distance is greater than 0. 6 wavelength, the eavesdropper key bit error rate is greater than 0. 48.
2911 在室外环境窃听距离大于1 倍波长后窃听方密钥误比特率为 0.47，实现了安全可靠的物理层密钥生成。 And in the outdoor environ-ment, after the eavesdropping distance is greater than 1 wavelength, the eavesdropper key bit error rate is 0. 47. Secure andreliable physical layer secret key generation is achieved.
2912 通过对频率抖动机理的研究，提出一种基于压控振荡器(Voltage-Controlled Oscillator，VCO)的真随机数发生器(True Random Number Generator，TRNG)设计方案。 After studies on the frequency jitter mechanism, a design of true random number generator(TRNG)basedon voltage-controlled oscillator(VCO)is proposed.