Semi-Supervised Cross-Subject Emotion Recognition Based on Stacked Denoising Autoencoder Architecture Using a Fusion of Multi-Modal Physiological Signals

In recent decades, emotion recognition has received considerable attention. As more enthusiasm has shifted to the physiological pattern, a wide range of elaborate physiological emotion data features come up and are combined with various classifying models to detect one’s emotional states. To circumvent the labor of artificially designing features, we propose to acquire affective and robust representations automatically through the Stacked Denoising Autoencoder (SDA) architecture with unsupervised pre-training, followed by supervised fine-tuning. In this paper, we compare the performances of different features and models through three binary classification tasks based on the Valence-Arousal-Dominance (VAD) affection model. Decision fusion and feature fusion of electroencephalogram (EEG) and peripheral signals are performed on hand-engineered features; data-level fusion is performed on deep-learning methods. It turns out that the fusion data perform better than the two modalities. To take advantage of deep-learning algorithms, we augment the original data and feed it directly into our training model. We use two deep architectures and another generative stacked semi-supervised architecture as references for comparison to test the method’s practical effects. The results reveal that our scheme slightly outperforms the other three deep feature extractors and surpasses the state-of-the-art of hand-engineered features.     

A Fault Detection Method for Electrohydraulic Switch Machine Based on Oil-Pressure-Signal-Sectionalized Feature Extraction

A turnout switch machine is key equipment in a railway, and its fault condition has an enormous impact on the safety of train operation. Electrohydraulic switch machines are increasingly used in high-speed railways, and how to extract effective fault features from their working condition monitoring signal is a difficult problem. This paper focuses on the sectionalized feature extraction method of the oil pressure signal of the electrohydraulic switch machine and realizes the fault detection of the switch machine based on this method. First, the oil pressure signal is divided into three stages according to the working principle and action process of the switch machine, and multiple features of each stage are extracted. Then the max-relevance and min-redundancy (mRMR) algorithm is applied to select the effective features. Finally, the mini batch k-means method is used to achieve unsupervised fault diagnosis. Through experimental verification, this method can not only derive the best sectionalization mode and feature types of the oil pressure signal, but also achieve the fault diagnosis and the prediction of the status of the electrohydraulic switch machine.     

在MapInfo中实现数字化地图"非监督分类"

利用MapBasic语言,引入遥感图象处理中"非监督分类"的概念,在MapInfo平台下将同一图层中的不同要素按照他们的颜色、线宽等特征信息进行分类提取,然后进行批量编辑.实现了数字化地形图中的"非监督分类",为实现大量的数据更新提供了简单、快速、智能的解决方案.提高了生产效率,在实际生产中得到了检验.     

Unsupervised cell identification on multidimensional X‐ray fluorescence datasets

A novel approach to locate, identify and refine positions and whole areas of cell structures based on elemental contents measured by X‐ray fluorescence microscopy is introduced. It is shown that, by initializing with only a handful of prototypical cell regions, this approach can obtain consistent identification of whole cells, even when cells are overlapping, without training by explicit annotation. It is robust both to different measurements on the same sample and to different initializations. This effort provides a versatile framework to identify targeted cellular structures from datasets too complex for manual analysis, like most X‐ray fluorescence microscopy data. Possible future extensions are also discussed.     

基于土壤系统分类的土壤遥感自动识别分类系统的设计

设计出基于土壤系统分类的土壤遥感自动识别分类系统(ASICS)的总体框架，并探讨了其实现方法．该系统由3部分组成：空间数据库、分类识别和结果输出．空间数据库由遥感数据和非遥感数据组成，遥感数据选择绿度指数图像、土壤专题信息增强图像和同谱土壤图像，非遥感数据包括高程、高差、坡度、坡向、粗糙率指数、水系密度、土壤发生分类类别．分类识别采取在非监督分类的基础上对土壤类型进行正向推理与逆向推理相结合的识别模式．用像结构建立了土壤分类识别的规则，构造了基于土壤系统分类的土壤分类判决树．     

基于残差卷积自注意力神经网络的铝电解过热度识别方法

过热度是反映铝电解槽当前生产效率的重要指标,由于过热度难以在线实时测量,本文提出一种基于残差卷积自注意力神经网络的过热度识别方法.针对铝电解生产过程数据为时间序列数据且具有多源异构特性,设计异构数据的同构表示方法.在此基础上建立残差卷积自注意力神经网络模型以提取同构时间序列数据的全局与局部特征.针对过热度数据标签少且类别分布不均匀问题,采用基于自动编码器的无监督预训练方法与加权交叉熵损失函数以提高过热度识别任务的性能.在基准数据集上进行仿真对比实验以验证本文所提方法的有效性,然后在只包含少量不平衡标签的铝电解过热度数据集上进行实验验证,结果表明本文构建的过热度识别模型相较与其他现有模型不仅提高了过热度识别准确率,而且在训练样本较少时保证了模型的泛化能力.     

ASTER高光谱影像在地面人工建筑物信息提取中的应用

ASTER(高级空间热辐射热反射探测仪)是高光谱遥感影像时代到来的标志,高光谱遥感与一般遥感影像相比能为对地观测提供更高质量的信息源.以福州市的一景ASTER影像数据进行主成分分析及波段运算,然后进行非监督自动分类,结果表明能较好地提取地面的人工建筑物信息.该结果可服务于城市规划和城市环境评价中.     

基于最小二乘法的无监督支持向量机

将最小二乘支持向量机的思想引入无监督学习,提出一个最小二乘无监督支持向量机.首先假设超平面过样本中心点,再给出线性可分的条件构造目标函数和约束条件,从而得到一个线性规划问题去求解聚类问题.     

无监督多尺度模糊聚类算法研究

针对模糊c均值聚类算法的一些不足之处提出了一种新的均值漂移聚类算法--无监督多尺度聚类算法.该算法不受初始化的影响,不用假定数据的聚类个数以及聚类中心的初始位置,能够利用模糊聚类的方法来获得硬的聚类划分,能够从不同的"划分尺度"揭示数据的聚类结构,并能自动的确定聚类个数.为了满足处理大数据集的需要,设计了快速无监督多尺度模糊聚类算法.通过实验证明无监督多尺度聚类算法在多数数据集上都表现良好且具有最好的总体聚类性能,并能成功揭示出数据的聚类结构.实验还证明快速无监督多尺度模糊聚类算法具有较快的速度和较高的识别精度且适用于大数据集.2个算法都取得了令人满意的实验结果.     

基于超像素与Span-Pauli分解的SAR影像分类

针对高分辨率极化合成孔径雷达(SAR)影像解译中面向像素方法难以充分利用影像信息的问题,提出一种基于超像素与Span-Pauli分解的非监督分类方法.利用分水岭方法易于过分割的特点,将分水岭分割得到的特征相似、空间相邻的像素集合视为超像素,并作为分类算法的基本处理单元.利用极化参数Span及Pauli基对极化SAR数据进行迭代分类,以实现适用于高分辨率SAR影像的非监督分类.采用X波段高分辨率SAR数据进行实验,结果表明:基于超像素的分类方法能有效抑制分类结果中的椒盐现象,将总体精度提高到了73.81%;在准确区分水体、道路的基础上,提高了复杂的农作物类别的分类精度.