| ID | 原文 | 译文 |
| 2473 | 及比较分析了它们在 UCF101 和 HMDB51 这两个数据集上的识别效果。 | Furthermore, the performances of some typical methods on UCF101 and HMDB51 datasetsare overviewed and analyzed. |
| 2474 | 最后分别从视频预处理、视频中人体运动信息表征、模型学习训练这三个角度对未来动作识别可能的发展方向进行了论述。 | Last the possible future research directions are discussed from three perspectives:the video datapreprocessing, the video human motion feature representation, and the model training. |
| 2475 | 本文提出一种基于混沌信号特性的信号盲提取算法，由于不同的混沌信号在相空间里面对应着不同的吸引子二阶增长率，利用这个特点定义了增殖系数(Proliferation Exponent，PE)并将其作为混沌信号提取的目标函数。 | In this paper, we propose a signal extraction algorithm based on the property of chaotic signal. Each chaoticsignal corresponds to a different chaotic attractor in phase space. We define proliferation exponent(PE)using the property a-bove;PE is used as a statistic feature to classify chaotic signal and computationally less dissipative compared with KullbackLeibler divergence. |
| 2476 | 首先分析基于增殖系数的梯度搜索方法在解决盲提取问题时存在不足，并将混沌信号的盲提取问题转化为带约束的优化问题， | We firstly model the problem of blind source extraction into an optimization problem with constrained. The objective function based on PE is non-convex or multi-model and solving the optimization problem with gradient searchmethod may lead to local optimum. |
| 2477 | 提出利用改进的粒子群优化算法解决信号盲提取的优化问题，通过惯性系数动态调整和最优位置的扰动，提高算法的寻优性能。 | We use particle swarm optimization(PSO)algorithm to solve the above optimizationproblem, the algorithm is improved by adjusting the inertia coefficient dynamically and the global optimal position is dis-turbed to diversify the particle population and increase the probability of escaping from local trap. |
| 2478 | 实验结果表明基于增殖系数的信号提取算法能有效地提取混沌信号，提取的信号在时域和相空间与源信号接近，同时算法也表现出对噪声污染的鲁棒性。 | The experimental results show that the proposed signal extraction algorithm can extract the mixture of chaotic signals and multi-channel Gaussian sig-nals efficiently. |
| 2479 | 针对现有的显著性检测算法检测目标类型单一、通用性差的问题，提出一种基于无监督栈式降噪自编码网络的显著性检测算法。 | The traditional saliency detection method is difficult to detect different kinds of saliency target simultane-ously. In order to solve this problem, an algorithm based on unsupervised SDAE network is proposed in this paper. |
| 2480 | 该算法利用无监督栈式降噪自编码网络(Stacked Denoising Auto Encoder，SDAE)在多个尺度对原始图像进行稀疏重构，将原始图像与 SDAE 网络重构图像之间的差作为显著图，二值化后的显著图作为显著性目标检测结果。 | The stacked denoising auto-encoder (SDAE)network is used to sparsely reconstruct original image in multiple scales. The difference between the original image and the reconstructed image is used as a saliency map, and the binaryzation of the sali-ency map is used as salient detection result. |
| 2481 | 在 SDAE 网络训练过程中，将原始图像作为原始数据，网络重构的图像作为观察数据。 | In the process of SDAE network training, the original image is used as the origi-nal data and the reconstructed images are treated as observed data. |
| 2482 | 为了提升网络训练效率，首先利用无监督逐层贪婪方法训练同结构的深度信念网络(Deep Belief Network，DBN)，将训练得到的 DBN 网络参数设为 SDAE 网络的初始参数，再计算原始数据与观察数据之间的互信息作为网络收敛代价，利用反向传播进行网络参数微调。 | In order to improve the efficiency of network training, thedeep belief network (DBN)is trained by greedy method in each layer without supervising, and the network parameters aredelivered to stacked denoising auto-encoder (SDAE)network as initial parameters. Then, the mutual information between the original data and the observed data is used as loss function, and the network parameters are tuned by back propagation. |