ID 原文 译文
19685 针对深度卷积网络原理分析的问题，该文提出一种基于模型重建的权值可视化方法。 A method for visualizing the weights of a reconstructed model is proposed to analyze a deep convolutional network works.
19686 首先利用原有的神经网络对测试样本进行前向传播，以获取重建模型所需要的先验信息； Firstly, a specific input is used in the original neural network during the forward propagation to get the prior information for model reconstruction.
19687 然后对原本网络中的部分结构进行修改，使其便于后续的参数计算； Then some of the structure of the originalnetwork is changed for further parameter calculation.
19688 再利用正交向量组，逐一地计算重建模型的参数； After that, the parameters of the reconstructed model arecalculated with a group of orthogonal vectors.
19689 最后将计算所得的参数按照特定的顺序进行重排列，实现权值的可视化。 Finally, the parameters are put into a special order to make them visualized.
19690 实验结果表明，对于满足一定条件的深度卷积网络，利用该文所提方法重建的模型在分类过程的前向传播运算中与原模型完全等效，并且可以明显观察到重建后模型的权值所具有的特征，从而分析神经网络实现图像分类的原理。 Experimental results show that the model reconstructed with the proposed method is totally equivalent to the original model during the forward propagation in the classification process. The feature of the weights of the reconstructed model can be observed clearly and the principle of the neural network can be analyzed with the feature.
19691 脑电信号一直被誉为疲劳检测的“金标准”，驾驶者的精神状态可通过对脑电信号的分析得到。 Electro Encephalo Gram (EEG) is regarded as a “gold standard” of fatigue detection and drivers’ vigilance states can be detected through the analysis of EEG signals.
19692 但由于脑电信号具有非线性、非平稳性和空间分辨率低等特点，传统的机器学习方法在运用脑电信号进行疲劳检测时还存在识别率低，特征提取操作繁琐等不足。 However, due to the characteristics of non-linear, non-stationary and low spatial resolution of EEG signals, traditional machine learning methods still have the disadvantages of low recognition rate and complicated feature extraction operations in EEG-based fatigue detection task.
19693 为此，该文基于脑电信号的电极-频率分布图，提出运用深度迁移学习实现的驾驶疲劳检测方法，即搭建深度卷积神经网络，并利用SEED脑电情绪数据集对其进行预训练，然后通过迁移学习方法将其用于驾驶疲劳检测。 To tackle this problem, a fatigue detection method with transfer learning based on theElectrode-Frequency Distribution Maps (EFDMs) of EEG signals is proposed. A deep convolutional neuralnetwork is designed and pre-trained with SEED dataset, and then it is used for fatigue detection with transfer learning strategy.
19694 实验结果表明，卷积神经网络模型能够很好地从电极-频率分布图中获得与疲劳状态相关的特征信息，达到较好的识别效果。 Experimental results show that the proposed convolutional neural network can automatically obtain vigilance related features from EFDMs, and achieve much better recognition results than traditional machine learning methods.