As the popularity of digital images is rapidly increasing on the Internet, research on technologies for semantic image classification has become an important research topic. However, the well-known content-based image...As the popularity of digital images is rapidly increasing on the Internet, research on technologies for semantic image classification has become an important research topic. However, the well-known content-based image classification methods do not overcome the so-called semantic gap problem in which low-level visual features cannot represent the high-level semantic content of images. Image classification using visual and textual information often performs poorly since the extracted textual features are often too limited to accurately represent the images. In this paper, we propose a semantic image classification ap- proach using multi-context analysis. For a given image, we model the relevant textual information as its multi-modal context, and regard the related images connected by hyperlinks as its link context. Two kinds of context analysis models, i.e., cross-modal correlation analysis and link-based correlation model, are used to capture the correlation among different modals of features and the topical dependency among images induced by the link structure. We propose a new collective classification model called relational support vector classifier (RSVC) based on the well-known Support Vector Machines (SVMs) and the link-based cor- relation model. Experiments showed that the proposed approach significantly improved classification accuracy over that of SVM classifiers using visual and/or textual features.展开更多
目前的脑电(EEG)情感识别模型忽略了不同时段情感状态的差异性,未能强化关键的情感信息。针对上述问题,提出一种多上下文向量优化的卷积递归神经网络(CR-MCV)。首先构造脑电信号的特征矩阵序列,通过卷积神经网络(CNN)学习多通道脑电的...目前的脑电(EEG)情感识别模型忽略了不同时段情感状态的差异性,未能强化关键的情感信息。针对上述问题,提出一种多上下文向量优化的卷积递归神经网络(CR-MCV)。首先构造脑电信号的特征矩阵序列,通过卷积神经网络(CNN)学习多通道脑电的空间特征;然后利用基于多头注意力的递归神经网络生成多上下文向量进行高层抽象特征提取;最后利用全连接层进行情感分类。在DEAP(Database for Emotion Analysis using Physiological signals)数据集上进行实验,CR-MCV在唤醒和效价维度上分类准确率分别为88.09%和89.30%。实验结果表明,CR-MCV在利用电极空间位置信息和不同时段情感状态显著性特征基础上,能够自适应地分配特征的注意力并强化情感状态显著性信息。展开更多
时间动作检测是视频理解领域中具有挑战性的任务。先前的时间动作检测模型主要关注视频帧的分类,而忽略视频帧之间的时序关系,导致时间动作检测模型的性能下降。为此,提出融合时序关系和上下文信息的时间动作检测方法(temporal action d...时间动作检测是视频理解领域中具有挑战性的任务。先前的时间动作检测模型主要关注视频帧的分类,而忽略视频帧之间的时序关系,导致时间动作检测模型的性能下降。为此,提出融合时序关系和上下文信息的时间动作检测方法(temporal action detection based on enhanced temporal relationship and context information,ETRD)。首先,设计了基于增强局部时序关系注意力机制的全局特征编码器,关注相邻帧的时序关系;其次,构建基于上下文信息的时序特征增强模块,融合上下文信息;最后,通过头部输出分类和回归结果。实验结果表明,所提出的ETRD模型在THUMOS14和ActivityNet1.3数据集上的平均mAP(mean average precision,平均精度均值)分别达到了67.5%和36.0%。相比于Actionformer模型的66.8%和35.6%,ETRD模型的平均mAP分别提升了0.7%和0.4%。利用视觉传感器,所提出的模型可检测出行为类别和持续时间。同时,结合心率等生理信号,可实现个体健康状态管理,为远程医疗、智能监控等提供了一种解决方案。展开更多
The problem of detecting community structures of a social network has been extensively studied over recent years, but most existing methods solely rely on the network structure and neglect the context information of t...The problem of detecting community structures of a social network has been extensively studied over recent years, but most existing methods solely rely on the network structure and neglect the context information of the social relations. The main reason is that a context-rich network offers too much flexibility and complexity for automatic or manual modulation of the multifaceted context in the analysis process. We address the challenging problem of incorporating context information into the community analysis with a novel visual analysis mechanism. Our approach consists of two stages： interactive discovery of salient context, and iterative context-guided community detection. Central to the analysis process is a context relevance model （CRM） that visually characterizes the influence of a given set of contexts on the variation of the detected communities, and discloses the community structure in specific context configurations. The extracted relevance is used to drive an iterative visual reasoning process, in which the community structures are progressively discovered. We introduce a suite of visual representations to encode the community structures, the context as well as the CRM. In particular, we propose an enhanced parallel coordinates representation to depict the context and community structures, which allows for interactive data exploration and community investigation. Case studies on several datasets demonstrate the efficiency and accuracy of our approach.展开更多
针对如何解决中文司法事件检测中触发词与上下文关系不足以判定事件实例、案件触发词表述相似以及同一个案件中多个触发词识别和分类模糊的问题,本研究提出一种基于多头指针的司法事件检测方法。首先,该方法将上下文信息和罪名特征融合...针对如何解决中文司法事件检测中触发词与上下文关系不足以判定事件实例、案件触发词表述相似以及同一个案件中多个触发词识别和分类模糊的问题,本研究提出一种基于多头指针的司法事件检测方法。首先,该方法将上下文信息和罪名特征融合作为输入,使用双向长短期记忆(Bi-directional Long Short-Term Memory,BiLSTM)网络捕获数据依赖关系,深入提取特征;然后,使用多头指针网络对字符间的依赖关系进行建模,有效捕捉句子中的触发词;最后,利用指针标注技术抽取触发词,实现司法事件的有效检测。在公开司法数据集LEVEN上实验验证该方法的有效性,其中微平均和宏平均的F1指标达到了87.53%和78.05%,优于现有模型。该方法不仅显著提高了事件触发词的识别精度,而且也增强了对复杂司法文本中事件上下文关系的把握能力。展开更多
基金Project supported by the Hi-Tech Research and Development Pro-gram (863) of China (No. 2003AA119010), and China-American Digital Academic Library (CADAL) Project (No. CADAL2004002)
文摘As the popularity of digital images is rapidly increasing on the Internet, research on technologies for semantic image classification has become an important research topic. However, the well-known content-based image classification methods do not overcome the so-called semantic gap problem in which low-level visual features cannot represent the high-level semantic content of images. Image classification using visual and textual information often performs poorly since the extracted textual features are often too limited to accurately represent the images. In this paper, we propose a semantic image classification ap- proach using multi-context analysis. For a given image, we model the relevant textual information as its multi-modal context, and regard the related images connected by hyperlinks as its link context. Two kinds of context analysis models, i.e., cross-modal correlation analysis and link-based correlation model, are used to capture the correlation among different modals of features and the topical dependency among images induced by the link structure. We propose a new collective classification model called relational support vector classifier (RSVC) based on the well-known Support Vector Machines (SVMs) and the link-based cor- relation model. Experiments showed that the proposed approach significantly improved classification accuracy over that of SVM classifiers using visual and/or textual features.
文摘识别非驾驶行为是提高驾驶安全性的重要手段之一。目前基于骨架序列和图像的融合识别方法具有计算量大和特征融合困难的问题。针对上述问题,本文提出一种基于多尺度骨架图和局部视觉上下文融合的驾驶员行为识别模型(skeleton-image based behavior recognition network,SIBBR-Net)。SIBBR-Net通过基于多尺度图的图卷积网络和基于局部视觉及注意力机制的卷积神经网络,充分提取运动和外观特征,较好地平衡了模型表征能力和计算量间的关系。基于手部运动的特征双向引导学习策略、自适应特征融合模块和静态特征空间上的辅助损失,使运动和外观特征间互相引导更新并实现自适应融合。最终在Drive&Act数据集进行算法测试,SIBBR-Net在动态标签和静态标签条件下的平均正确率分别为61.78%和80.42%,每秒浮点运算次数为25.92G,较最优方法降低了76.96%。
文摘目前的脑电(EEG)情感识别模型忽略了不同时段情感状态的差异性,未能强化关键的情感信息。针对上述问题,提出一种多上下文向量优化的卷积递归神经网络(CR-MCV)。首先构造脑电信号的特征矩阵序列,通过卷积神经网络(CNN)学习多通道脑电的空间特征;然后利用基于多头注意力的递归神经网络生成多上下文向量进行高层抽象特征提取;最后利用全连接层进行情感分类。在DEAP(Database for Emotion Analysis using Physiological signals)数据集上进行实验,CR-MCV在唤醒和效价维度上分类准确率分别为88.09%和89.30%。实验结果表明,CR-MCV在利用电极空间位置信息和不同时段情感状态显著性特征基础上,能够自适应地分配特征的注意力并强化情感状态显著性信息。
文摘时间动作检测是视频理解领域中具有挑战性的任务。先前的时间动作检测模型主要关注视频帧的分类,而忽略视频帧之间的时序关系,导致时间动作检测模型的性能下降。为此,提出融合时序关系和上下文信息的时间动作检测方法(temporal action detection based on enhanced temporal relationship and context information,ETRD)。首先,设计了基于增强局部时序关系注意力机制的全局特征编码器,关注相邻帧的时序关系;其次,构建基于上下文信息的时序特征增强模块,融合上下文信息;最后,通过头部输出分类和回归结果。实验结果表明,所提出的ETRD模型在THUMOS14和ActivityNet1.3数据集上的平均mAP(mean average precision,平均精度均值)分别达到了67.5%和36.0%。相比于Actionformer模型的66.8%和35.6%,ETRD模型的平均mAP分别提升了0.7%和0.4%。利用视觉传感器,所提出的模型可检测出行为类别和持续时间。同时,结合心率等生理信号,可实现个体健康状态管理,为远程医疗、智能监控等提供了一种解决方案。
基金supported by the National Natural Science Foundation of China under Grant Nos. 61232012, 61202279the National High Technology Research and Development 863 Program of China under Grant No. 2012AA12090+1 种基金the Natural Science Foundation of Zhejiang Province of China under Grant No. LR13F020001the Doctoral Fund of Ministry of Education of China under Grant No. 20120101110134
文摘The problem of detecting community structures of a social network has been extensively studied over recent years, but most existing methods solely rely on the network structure and neglect the context information of the social relations. The main reason is that a context-rich network offers too much flexibility and complexity for automatic or manual modulation of the multifaceted context in the analysis process. We address the challenging problem of incorporating context information into the community analysis with a novel visual analysis mechanism. Our approach consists of two stages： interactive discovery of salient context, and iterative context-guided community detection. Central to the analysis process is a context relevance model （CRM） that visually characterizes the influence of a given set of contexts on the variation of the detected communities, and discloses the community structure in specific context configurations. The extracted relevance is used to drive an iterative visual reasoning process, in which the community structures are progressively discovered. We introduce a suite of visual representations to encode the community structures, the context as well as the CRM. In particular, we propose an enhanced parallel coordinates representation to depict the context and community structures, which allows for interactive data exploration and community investigation. Case studies on several datasets demonstrate the efficiency and accuracy of our approach.
文摘针对如何解决中文司法事件检测中触发词与上下文关系不足以判定事件实例、案件触发词表述相似以及同一个案件中多个触发词识别和分类模糊的问题,本研究提出一种基于多头指针的司法事件检测方法。首先,该方法将上下文信息和罪名特征融合作为输入,使用双向长短期记忆(Bi-directional Long Short-Term Memory,BiLSTM)网络捕获数据依赖关系,深入提取特征;然后,使用多头指针网络对字符间的依赖关系进行建模,有效捕捉句子中的触发词;最后,利用指针标注技术抽取触发词,实现司法事件的有效检测。在公开司法数据集LEVEN上实验验证该方法的有效性,其中微平均和宏平均的F1指标达到了87.53%和78.05%,优于现有模型。该方法不仅显著提高了事件触发词的识别精度,而且也增强了对复杂司法文本中事件上下文关系的把握能力。
