ID 原文 译文
17525 首先根据交叠信号量测的特征值分布来确定交叠信号的个数； Firstly, the number of overlappingsignals is determined with the eigenvalues distribution of the measurements.
17526 然后利用MUSIC算法作谱峰搜索得到各信号的DOA，并重构混合矩阵； Secondly, the mixing matrix withthe DOA of signals, which is estimated by peak value searching in MUSIC algorithm.
17527 最后通过计算混合矩阵的广义逆得到分离矩阵，并实现对交叠信号的分离。 Finally, the separatingmatrix is estimated by calculating the Moore-Penrose inverse of the reconstructed mixing matrix, achievingseparation of overlapping signals.
17528 以6阵元均匀线阵为前提进行仿真分析，结果表明所提分离算法可达到90%以上的分离成功率，分离性能和独立成分分析(ICA)算法相当，优于基于投影技术分离算法(PA)，但计算量远小于ICA算法，不足ICA算法计算量1/10，更易于工程化应用。 Simulation is done based on uniform linear array with 6 elements. The resultsshow that the proposed separation algorithm can achieve more than 90% success rate to separate two short ModeS replies, and the separating performance is similar to the Independent Component Analysis (ICA) algorithmand is better than Projection Algorithm (PA). The amount of calculation is less than 10 percent of ICAalgorithm, thus the proposed separation algorithm is easier to engineering application.
17529 行人重识别的关键依赖于行人特征的提取，卷积神经网络具有强大的特征提取以及表达能力。 The key to person re-identification depends on the extraction of pedestrian characteristics.Convolutional neural networks have powerful feature extraction and expression capabilities.
17530 针对不同尺度下可以观察到不同的特征，该文提出一种基于多尺度和注意力网络融合的行人重识别方法(MSAN)。 In view of the factthat different features can be observed at different scales, a pedestrian re-identification method based on Multi-Scale Attention Network(MSAN) fusion is proposed.
17531 该方法通过对网络不同深度的特征进行采样，将采样的特征融合后对行人进行预测。 This method samples the features at different depths ofthe network and fuses the sampled features to predict pedestrians.
17532 不同深度的特征图具有不同的表达能力，使网络可以学习到行人身上更加细粒度的特征。 Feature maps of different depths havedifferent expressive powers, enabling the network to learn more fine-grained features of pedestrians.
17533 同时将注意力模块嵌入到残差网络中，使得网络能更加关注于一些关键信息，增强网络特征学习能力。 At the same time, the attention module is embedded in the residual network, so that the network can pay more attention to some key information and enhance the network feature learning ability.
17534 所提方法在Market1501, DukeMTMC-reID和MSMT17_V1数据集上首位准确率分别到了95.3%, 89.8%和82.2%。 The accuracy of the proposed method on the datasets such as Market1501, DukeMTMC-reID and MSMT17_V1 reaches 95.3%,89.8% and 82.2%, respectively.