基于改进的Random Subspace的客户投诉分类方法

电信业的客户投诉不断增多而又亟待高效处理。针对电信客户投诉数据的特点，提出了一种面向高维数据的改进的集成学习分类方法。该方法综合考虑客户投诉中的文本信息及客户通讯状态信息，基于Random Subspace方法，以支持向量机（Support Vector Machine，SVM）为基分类器，采用证据推理（Evidential Reasoning，ER）规则为一种新的集成策略，构造分类模型对电信客户投诉进行分类。所提模型和方法在某电信公司客户投诉数据上进行了验证，实验结果显示该方法能够显著提高客户投诉分类的准确率和投诉处理效率。     

不同照度下煤矸图像灰度及纹理特征提取的实验研究

研究煤和矸石在不同照度下的响应特性，设计了差异照度煤矸图像采集系统，进行了煤矸图像采集实验，建立了煤矸灰度子图像数据库，对数据库中的子图像的灰度、纹理特征进行了提取，通过定义归一化特征差异指数以及基于支持向量机(SVM)对实验数据进行了分析与讨论。结果表明:在同一照度下，煤和矸石在灰度及纹理特征上存在差异，且随着照度的改变，煤和矸石的特征也发生变化，同时两者的变化规律存在显著差异。基于不同特征的SVM分类器的识别正确率不同，照度的变化也会对分类器的识别正确率产生影响;当考虑照度因素后，分类器的识别正确率最大增加了13.31%，此外基于多特征多照度融合的SVM分类器性能较好，识别正确率为98-39%。     

Human action recognition toward massive-scale sport sceneries based on deep multi-model feature fusion

In sport sceneries, automatically recognizing human actions is a useful technique that can be popularly applied in may domains, such as human body tracking and athlete behavior analysis Most state-of-the-art deep architectures have achieved competitive performance in recognizing human action. However, it is still a challenging task due to the unavoidable occlusion, camera angle changes, and varied human posture. In this paper, we propose a novel deep multimodal feature fusion algorithm for human action recognition. The key technique is a multi-model feature fusion scheme. More specifically, we fuse visual feature, skeleton posture, probability maps and audio signal into a hybrid feature, which is utilized to represent human action. Then these feature channels are optimally combined using a deep model, wherein the weights of multiple feature channels can be predicted intelligently. Finally, the optimally fused feature are fed into a multi-class SVM for conducting human action recognition. Extensive comparative results and parameter analysis have shown the effectiveness of our proposed method.     

结合深度学习与支持向量机的金属零件识别

目的 在视觉引导的工业机器人自动拾取研究中，关键技术难点之一是机器人抓取目标区域的识别问题。特别是金属零件，其表面的反光、随意摆放时相互遮挡等非结构化因素都给抓取区域的识别带来巨大的挑战。因此，本文提出一种结合深度学习和支持向量机的抓取区域识别方法。方法 分别提取抓取区域的方向梯度直方图（HOG）和局部二进制模式（LBP）特征，利用主成分分析法（PCA）对融合后的特征进行降维，以此来训练支持向量机（SVM）分类器。通过训练Mask R-CNN（regions with convolutional neural network）神经网络完成抓取区域的初步分割。然后利用SVM对Mask R-CNN识别的抓取区域进行二次分类，完成对干扰区域的剔除。最后计算掩码完成实例分割，以此达到对抓取区域的精确识别。结果 对于随机摆放的铜质金属零件，本文算法与单一的Mask R-CNN及多特征融合的SVM算法就识别准确率、错检率、漏检率3个指标进行了比较，结果表明本文算法在识别准确率上较Mask R-CNN和SVM算法分别提高了7%和25%，同时有效降低了错检率与漏检率。结论 本文算法结合了Mask R-CNN与SVM两种方法，对于反光和遮挡情况具有一定的鲁棒性，同时有效地提升了目标识别的准确率。     

Efficient fault diagnosis method of PEMFC thermal management system for various current densities

The temperature of a fuel cell has a considerable impact on the saturation of a membrane, electrochemical reaction speed, and durability. So thermal management is considered one of the critical issues in polymer electrolyte membrane fuel cells. Therefore, the reliability of the thermal management system is also crucial for the performance and durability of a fuel cell system. In this work, a methodology for component-level fault diagnosis of polymer electrolyte membrane fuel cell thermal management system for various current densities is proposed. Specifically, this study suggests fault diagnosis using limited data, based on an experimental approach. Normal and five component-level fault states are diagnosed with a support vector machine model using temperature, pressure, and fan control signal data. The effects of training data at different operating current densities on fault diagnosis are analyzed. The effects of data preprocessing method are investigated, and the cause of misdiagnosis is analyzed. On this basis, diagnosis results show that the proposed methodology can realize efficient component-level fault diagnosis using limited data. The diagnosis accuracy is over 92% when the residual basis scaling method is used, and data at the highest operating current density is used to train the support vector machine.     

采用信息增益率的混合入侵检测模型设计

针对现有混合入侵检测模型仅定性选取特征而导致检测精度较低的问题，同时为了充分结合误用检测模型和异常检测模型的优势，提出一种采用信息增益率的混合入侵检测模型.首先，利用信息增益率定量地选择特征子集，最大程度地保留样本信息；其次，采用余弦时变粒子群算法确定支持向量机参数构建误用检测模型，使其更好地平衡粒子在全局和局部的搜索能力，然后，选取灰狼算法确定单类支持向量机参数构建异常检测模型，以此来提高对最优参数的搜索效率和精细程度，综合提高混合入侵检测模型对攻击的检测效果；最后，通过两种数据集进行仿真实验，验证了所提混合入侵检测模型具有较好的检测性能.     

栈式降噪自编码器在波形单元识别中的应用

为提高多功能雷达(Multi-Function Radar,MFR)波形单元的识别准确率和鲁棒性,提出一种栈式降噪自编码器(Stacked Denoising Autoencoders,SDAE)与支持向量机(Support Vector Machine,SVM)相结合的波形单元识别方法.首先摒弃传统依赖脉冲序列分析技术对MFR信号进行处理的方法,通过分析波形单元结构并借助参数间的联合变化特征,提出一种MFR波形单元分段识别模型,将传统对脉冲序列的识别转化为对MFR波形单元的识别;然后在该模型的基础上引入SDAE,对训练样本数据、SDAE隐含层神经元节点进行加噪处理,并利用这些加噪后的样本数据训练优化SDAE网络模型,提取出样本数据的深层稳健特征;最后引入SVM算法,借助SDAE挖掘出的样本深层特征,实现SVM模型的优化,得到最终的波形单元识别模型(SDAE-SVM).仿真实验表明:提出的波形单元识别方法在相同样本数目和测试误差的条件下,与SVM算法相比,能够取得较高的识别准确率,具备更优越的识别效果.证实MFR波形单元识别模型是有效的,且通过SDAE网络的引入,使得SDAE-SVM方法能够自主地挖掘原始信号的深层特征,提高波形单元识别的鲁棒性和准确率.     

基于太赫兹时域光谱的芝麻油品种识别研究

为了实现对芝麻油品种的快速鉴别,本文基于太赫兹时域光谱(THz-TDS)提出一种芝麻油品种识别的方法。选取0~2.5 THz范围内的光谱进行分析,通过主成分分析法(PCA)对时域光谱数据进行降维,选择前4个主成分(累计贡献率大于99%)代表原始数据,然后利用支持向量机(SVM)方法对不同品种芝麻油进行分类识别,分类时使用3种不同的核函数建模,并采用网格搜索算法获得最优模型及其模型参数。使用径向基核函数(参数为惩罚函数C=0.01,核函数系数γ=0.1)的模型识别率最高,达到100%,说明太赫兹时域光谱技术结合PCA和SVM方法可以快速可靠的进行食用油的识别,为食品安全的识别提供一种新的技术手段。     

Robust SVM-direct torque control of induction motor based on sliding mode controller and sliding mode observer

This paper proposes a design of control and estimation strategy for induction motor based on the variable structure approach. It describes a coupling of sliding mode direct torque control (DTC) with sliding mode flux and speed observer. This algorithm uses direct torque control basics and the sliding mode approach. A robust electromagnetic torque and flux controllers are designed to overcome the conventional SVM-DTC drawbacks and to ensure fast response and full reference tracking with desired dynamic behavior and low ripple level. The sliding mode controller is used to generate reference voltages in stationary frame and give them to the controlled motor after modulation by a space vector modulation (SVM) inverter. The second aim of this paper is to design a sliding mode speed/flux observer which can improve the control performances by using a sensorless algorithm to get an accurate estimation, and consequently, increase the reliability of the system and decrease the cost of using sensors. The effectiveness of the whole composed control algorithm is investigated in different robustness tests with simulation using Matlab/Simulink and verified by real time experimental implementation based on dS pace 1104 board.     

Structural Aspects and Prediction of Calmodulin-Binding Proteins

Calmodulin (CaM) is an important intracellular protein that binds Ca²⁺ and functions as a critical second messenger involved in numerous biological activities through extensive interactions with proteins and peptides. CaM’s ability to adapt to binding targets with different structures is related to the flexible central helix separating the N- and C-terminal lobes, which allows for conformational changes between extended and collapsed forms of the protein. CaM-binding targets are most often identified using prediction algorithms that utilize sequence and structural data to predict regions of peptides and proteins that can interact with CaM. In this review, we provide an overview of different CaM-binding proteins, the motifs through which they interact with CaM, and shared properties that make them good binding partners for CaM. Additionally, we discuss the historical and current methods for predicting CaM binding, and the similarities and differences between these methods and their relative success at prediction. As new CaM-binding proteins are identified and classified, we will gain a broader understanding of the biological processes regulated through changes in Ca²⁺ concentration through interactions with CaM.