Background The use of micro-expression recognition to recognize human emotions is one of the most critical challenges in human-computer interaction applications. In recent years, cross-database micro-expression recogn...Background The use of micro-expression recognition to recognize human emotions is one of the most critical challenges in human-computer interaction applications. In recent years, cross-database micro-expression recognition(CDMER) has emerged as a significant challenge in micro-expression recognition and analysis. Because the training and testing data in CDMER come from different micro-expression databases, CDMER is more challenging than conventional micro-expression recognition. Methods In this paper, an adaptive spatio-temporal attention neural network(ASTANN) using an attention mechanism is presented to address this challenge. To this end, the micro-expression databases SMIC and CASME II are first preprocessed using an optical flow approach,which extracts motion information among video frames that represent discriminative features of micro-expression.After preprocessing, a novel adaptive framework with a spatiotemporal attention module was designed to assign spatial and temporal weights to enhance the most discriminative features. The deep neural network then extracts the cross-domain feature, in which the second-order statistics of the sample features in the source domain are aligned with those in the target domain by minimizing the correlation alignment(CORAL) loss such that the source and target databases share similar distributions. Results To evaluate the performance of ASTANN, experiments were conducted based on the SMIC and CASME II databases under the standard experimental evaluation protocol of CDMER. The experimental results demonstrate that ASTANN outperformed other methods in relevant crossdatabase tasks. Conclusions Extensive experiments were conducted on benchmark tasks, and the results show that ASTANN has superior performance compared with other approaches. This demonstrates the superiority of our method in solving the CDMER problem.展开更多
In Unsupervised Domain Adaptation(UDA)for person re-identification(re-ID),the primary challenge is reducing the distribution discrepancy between the source and target domains.This can be achieved by implicitly or expl...In Unsupervised Domain Adaptation(UDA)for person re-identification(re-ID),the primary challenge is reducing the distribution discrepancy between the source and target domains.This can be achieved by implicitly or explicitly constructing an appropriate intermediate domain to enhance recognition capability on the target domain.Implicit construction is difficult due to the absence of intermediate state supervision,making smooth knowledge transfer from the source to the target domain a challenge.To explicitly construct the most suitable intermediate domain for the model to gradually adapt to the feature distribution changes from the source to the target domain,we propose the Minimal Transfer Cost Framework(MTCF).MTCF considers all scenarios of the intermediate domain during the transfer process,ensuring smoother and more efficient domain alignment.Our framework mainly includes threemodules:Intermediate Domain Generator(IDG),Cross-domain Feature Constraint Module(CFCM),and Residual Channel Space Module(RCSM).First,the IDG Module is introduced to generate all possible intermediate domains,ensuring a smooth transition of knowledge fromthe source to the target domain.To reduce the cross-domain feature distribution discrepancy,we propose the CFCM Module,which quantifies the difficulty of knowledge transfer and ensures the diversity of intermediate domain features and their semantic relevance,achieving alignment between the source and target domains by incorporating mutual information and maximum mean discrepancy.We also design the RCSM,which utilizes attention mechanism to enhance the model’s focus on personnel features in low-resolution images,improving the accuracy and efficiency of person re-ID.Our proposed method outperforms existing technologies in all common UDA re-ID tasks and improves the Mean Average Precision(mAP)by 2.3%in the Market to Duke task compared to the state-of-the-art(SOTA)methods.展开更多
为提高新能源汽车领域术语抽取准确率,面向新能源汽车专利文本提出一种领域术语抽取模型。传统的领域术语抽取方法过度依赖人工定义特征和领域知识,无法自动挖掘隐含特征,其识别性能过度依赖所选特征的质量。从深度学习的角度出发,提出...为提高新能源汽车领域术语抽取准确率,面向新能源汽车专利文本提出一种领域术语抽取模型。传统的领域术语抽取方法过度依赖人工定义特征和领域知识,无法自动挖掘隐含特征,其识别性能过度依赖所选特征的质量。从深度学习的角度出发,提出了一种基于attention的双向长短时记忆网络(bidirectional long short-term memory,BLSTM)与条件随机场(conditional random fields,CRF)相结合的领域术语抽取模型(BLSTM_attention_CRF模型),并使用基于词典与规则相结合的方法对结果进行校正,准确率可达到86%以上,方法切实可行。展开更多
现有跨域人脸活体检测算法,其特征提取过程容易发生过拟合和缺乏特征聚合所导致的泛化性不足问题。针对该问题,提出了跨域人脸活体检测的单边对抗网络算法,将分组卷积与改进的倒残差结构融合替换普通卷积,降低网络参数同时加强人脸细粒...现有跨域人脸活体检测算法,其特征提取过程容易发生过拟合和缺乏特征聚合所导致的泛化性不足问题。针对该问题,提出了跨域人脸活体检测的单边对抗网络算法,将分组卷积与改进的倒残差结构融合替换普通卷积,降低网络参数同时加强人脸细粒度特征的表达能力,并引入自适应特征归一化模块,强调图像中人脸活体信息区域淡化无关背景区域,有效避免人脸活体信息的过拟合并加强来自不同源域的人脸活体检测能力。基于NetVLAD引入通道注意力机制模块,通道注意力机制模块作为特征聚合网络的分支,学习不同源域中人脸局部特征的语义信息,有效增强对不同源域的人脸活体信息分类的泛化能力。设计两模块融合网络以提高未知场景下跨域人脸活体检测精度。在OULU-NPU、CASIA-FASD、MSU-MFSD和Idiap Replay-Attack数据集上的实验结果表明,该算法在跨数据集测试O&C&M to I、O&C&I to M、I&C&M to O、O&M&I to C均有不错的表现,其中,在O&C&I to M及O&M&I to C性能评估指标分别提升了0.99个百分点和0.5个百分点的精度。展开更多
Command and control(C2)servers are used by attackers to operate communications.To perform attacks,attackers usually employee the Domain Generation Algorithm(DGA),with which to confirm rendezvous points to their C2 ser...Command and control(C2)servers are used by attackers to operate communications.To perform attacks,attackers usually employee the Domain Generation Algorithm(DGA),with which to confirm rendezvous points to their C2 servers by generating various network locations.The detection of DGA domain names is one of the important technologies for command and control communication detection.Considering the randomness of the DGA domain names,recent research in DGA detection applyed machine learning methods based on features extracting and deep learning architectures to classify domain names.However,these methods are insufficient to handle wordlist-based DGA threats,which generate domain names by randomly concatenating dictionary words according to a special set of rules.In this paper,we proposed a a deep learning framework ATT-CNN-BiLSTMfor identifying and detecting DGA domains to alleviate the threat.Firstly,the Convolutional Neural Network(CNN)and bidirectional Long Short-Term Memory(BiLSTM)neural network layer was used to extract the features of the domain sequences information;secondly,the attention layer was used to allocate the corresponding weight of the extracted deep information from the domain names.Finally,the different weights of features in domain names were put into the output layer to complete the tasks of detection and classification.Our extensive experimental results demonstrate the effectiveness of the proposed model,both on regular DGA domains and DGA that hard to detect such as wordlist-based and part-wordlist-based ones.To be precise,we got a F1 score of 98.79%for the detection and macro average precision and recall of 83%for the classification task of DGA domain names.展开更多
现有方面级情感分析研究大多数往往从文本数据本身进行情感分析,而没有充分利用领域知识,忽略了语义依存信息的重要性,使得方面表示受噪声信息影响严重,出现噪声词注意权重高的可能。针对以上问题,结合领域知识,提出了一种剪枝算法和语...现有方面级情感分析研究大多数往往从文本数据本身进行情感分析,而没有充分利用领域知识,忽略了语义依存信息的重要性,使得方面表示受噪声信息影响严重,出现噪声词注意权重高的可能。针对以上问题,结合领域知识,提出了一种剪枝算法和语义-注意力机制相结合的方法(Pruning And Semantic At tention,PASA)针对服务领域特定方面进行情感分类。方法一方面结合领域知识对文本对应的语义依存树进行剪枝实现方面信息降噪,另一方面,通过利用语义-注意力机制进行增强并精确捕获方面的上下文描述信息,从而实现对方面情感极性的判断。为了验证所提出方法的正确性和有效性,在物流数据集、酒店评论数据集及SemEval 2014的Restaurant数据集进行了大量实验,结果表明,所提出的方法相对于其它方法具有明显优势,在垂直领域具有较好的应用前景。展开更多
Command and control(C2)servers are used by attackers to operate communications.To perform attacks,attackers usually employee the Domain Generation Algorithm(DGA),with which to confirm rendezvous points to their C2 ser...Command and control(C2)servers are used by attackers to operate communications.To perform attacks,attackers usually employee the Domain Generation Algorithm(DGA),with which to confirm rendezvous points to their C2 servers by generating various network locations.The detection of DGA domain names is one of the important technologies for command and control communication detection.Considering the randomness of the DGA domain names,recent research in DGA detection applyed machine learning methods based on features extracting and deep learning architectures to classify domain names.However,these methods are insufficient to handle wordlist-based DGA threats,which generate domain names by randomly concatenating dictionary words according to a special set of rules.In this paper,we proposed a a deep learning framework ATT-CNN-BiLSTMfor identifying and detecting DGA domains to alleviate the threat.Firstly,the Convolutional Neural Network(CNN)and bidirectional Long Short-Term Memory(BiLSTM)neural network layer was used to extract the features of the domain sequences information;secondly,the attention layer was used to allocate the corresponding weight of the extracted deep information from the domain names.Finally,the different weights of features in domain names were put into the output layer to complete the tasks of detection and classification.Our extensive experimental results demonstrate the effectiveness of the proposed model,both on regular DGA domains and DGA that hard to detect such as wordlist-based and part-wordlist-based ones.To be precise,we got a F1 score of 98.79% for the detection and macro average precision and recall of 83% for the classification task of DGA domain names.展开更多
文摘Background The use of micro-expression recognition to recognize human emotions is one of the most critical challenges in human-computer interaction applications. In recent years, cross-database micro-expression recognition(CDMER) has emerged as a significant challenge in micro-expression recognition and analysis. Because the training and testing data in CDMER come from different micro-expression databases, CDMER is more challenging than conventional micro-expression recognition. Methods In this paper, an adaptive spatio-temporal attention neural network(ASTANN) using an attention mechanism is presented to address this challenge. To this end, the micro-expression databases SMIC and CASME II are first preprocessed using an optical flow approach,which extracts motion information among video frames that represent discriminative features of micro-expression.After preprocessing, a novel adaptive framework with a spatiotemporal attention module was designed to assign spatial and temporal weights to enhance the most discriminative features. The deep neural network then extracts the cross-domain feature, in which the second-order statistics of the sample features in the source domain are aligned with those in the target domain by minimizing the correlation alignment(CORAL) loss such that the source and target databases share similar distributions. Results To evaluate the performance of ASTANN, experiments were conducted based on the SMIC and CASME II databases under the standard experimental evaluation protocol of CDMER. The experimental results demonstrate that ASTANN outperformed other methods in relevant crossdatabase tasks. Conclusions Extensive experiments were conducted on benchmark tasks, and the results show that ASTANN has superior performance compared with other approaches. This demonstrates the superiority of our method in solving the CDMER problem.
文摘In Unsupervised Domain Adaptation(UDA)for person re-identification(re-ID),the primary challenge is reducing the distribution discrepancy between the source and target domains.This can be achieved by implicitly or explicitly constructing an appropriate intermediate domain to enhance recognition capability on the target domain.Implicit construction is difficult due to the absence of intermediate state supervision,making smooth knowledge transfer from the source to the target domain a challenge.To explicitly construct the most suitable intermediate domain for the model to gradually adapt to the feature distribution changes from the source to the target domain,we propose the Minimal Transfer Cost Framework(MTCF).MTCF considers all scenarios of the intermediate domain during the transfer process,ensuring smoother and more efficient domain alignment.Our framework mainly includes threemodules:Intermediate Domain Generator(IDG),Cross-domain Feature Constraint Module(CFCM),and Residual Channel Space Module(RCSM).First,the IDG Module is introduced to generate all possible intermediate domains,ensuring a smooth transition of knowledge fromthe source to the target domain.To reduce the cross-domain feature distribution discrepancy,we propose the CFCM Module,which quantifies the difficulty of knowledge transfer and ensures the diversity of intermediate domain features and their semantic relevance,achieving alignment between the source and target domains by incorporating mutual information and maximum mean discrepancy.We also design the RCSM,which utilizes attention mechanism to enhance the model’s focus on personnel features in low-resolution images,improving the accuracy and efficiency of person re-ID.Our proposed method outperforms existing technologies in all common UDA re-ID tasks and improves the Mean Average Precision(mAP)by 2.3%in the Market to Duke task compared to the state-of-the-art(SOTA)methods.
文摘为提高新能源汽车领域术语抽取准确率,面向新能源汽车专利文本提出一种领域术语抽取模型。传统的领域术语抽取方法过度依赖人工定义特征和领域知识,无法自动挖掘隐含特征,其识别性能过度依赖所选特征的质量。从深度学习的角度出发,提出了一种基于attention的双向长短时记忆网络(bidirectional long short-term memory,BLSTM)与条件随机场(conditional random fields,CRF)相结合的领域术语抽取模型(BLSTM_attention_CRF模型),并使用基于词典与规则相结合的方法对结果进行校正,准确率可达到86%以上,方法切实可行。
文摘现有跨域人脸活体检测算法,其特征提取过程容易发生过拟合和缺乏特征聚合所导致的泛化性不足问题。针对该问题,提出了跨域人脸活体检测的单边对抗网络算法,将分组卷积与改进的倒残差结构融合替换普通卷积,降低网络参数同时加强人脸细粒度特征的表达能力,并引入自适应特征归一化模块,强调图像中人脸活体信息区域淡化无关背景区域,有效避免人脸活体信息的过拟合并加强来自不同源域的人脸活体检测能力。基于NetVLAD引入通道注意力机制模块,通道注意力机制模块作为特征聚合网络的分支,学习不同源域中人脸局部特征的语义信息,有效增强对不同源域的人脸活体信息分类的泛化能力。设计两模块融合网络以提高未知场景下跨域人脸活体检测精度。在OULU-NPU、CASIA-FASD、MSU-MFSD和Idiap Replay-Attack数据集上的实验结果表明,该算法在跨数据集测试O&C&M to I、O&C&I to M、I&C&M to O、O&M&I to C均有不错的表现,其中,在O&C&I to M及O&M&I to C性能评估指标分别提升了0.99个百分点和0.5个百分点的精度。
基金Our research was supported by the National Key Research and Development Program of China(Grant No.2016YFB0801004)the Strategic Priority Research Program of Chinese Academy of Sciences(Grant No.XDC02030200)the National Key Research and Development Program of China(Grant No.2018YFC0824801).
文摘Command and control(C2)servers are used by attackers to operate communications.To perform attacks,attackers usually employee the Domain Generation Algorithm(DGA),with which to confirm rendezvous points to their C2 servers by generating various network locations.The detection of DGA domain names is one of the important technologies for command and control communication detection.Considering the randomness of the DGA domain names,recent research in DGA detection applyed machine learning methods based on features extracting and deep learning architectures to classify domain names.However,these methods are insufficient to handle wordlist-based DGA threats,which generate domain names by randomly concatenating dictionary words according to a special set of rules.In this paper,we proposed a a deep learning framework ATT-CNN-BiLSTMfor identifying and detecting DGA domains to alleviate the threat.Firstly,the Convolutional Neural Network(CNN)and bidirectional Long Short-Term Memory(BiLSTM)neural network layer was used to extract the features of the domain sequences information;secondly,the attention layer was used to allocate the corresponding weight of the extracted deep information from the domain names.Finally,the different weights of features in domain names were put into the output layer to complete the tasks of detection and classification.Our extensive experimental results demonstrate the effectiveness of the proposed model,both on regular DGA domains and DGA that hard to detect such as wordlist-based and part-wordlist-based ones.To be precise,we got a F1 score of 98.79%for the detection and macro average precision and recall of 83%for the classification task of DGA domain names.
文摘现有方面级情感分析研究大多数往往从文本数据本身进行情感分析,而没有充分利用领域知识,忽略了语义依存信息的重要性,使得方面表示受噪声信息影响严重,出现噪声词注意权重高的可能。针对以上问题,结合领域知识,提出了一种剪枝算法和语义-注意力机制相结合的方法(Pruning And Semantic At tention,PASA)针对服务领域特定方面进行情感分类。方法一方面结合领域知识对文本对应的语义依存树进行剪枝实现方面信息降噪,另一方面,通过利用语义-注意力机制进行增强并精确捕获方面的上下文描述信息,从而实现对方面情感极性的判断。为了验证所提出方法的正确性和有效性,在物流数据集、酒店评论数据集及SemEval 2014的Restaurant数据集进行了大量实验,结果表明,所提出的方法相对于其它方法具有明显优势,在垂直领域具有较好的应用前景。
基金supported by the National Key Research and Development Program of China(Grant No.2016YFB0801004)the Strategic Priority Research Program of Chinese Academy of Sciences(Grant No.XDC02030200)the National Key Research and Development Program of China(Grant No.2018YFC0824801).
文摘Command and control(C2)servers are used by attackers to operate communications.To perform attacks,attackers usually employee the Domain Generation Algorithm(DGA),with which to confirm rendezvous points to their C2 servers by generating various network locations.The detection of DGA domain names is one of the important technologies for command and control communication detection.Considering the randomness of the DGA domain names,recent research in DGA detection applyed machine learning methods based on features extracting and deep learning architectures to classify domain names.However,these methods are insufficient to handle wordlist-based DGA threats,which generate domain names by randomly concatenating dictionary words according to a special set of rules.In this paper,we proposed a a deep learning framework ATT-CNN-BiLSTMfor identifying and detecting DGA domains to alleviate the threat.Firstly,the Convolutional Neural Network(CNN)and bidirectional Long Short-Term Memory(BiLSTM)neural network layer was used to extract the features of the domain sequences information;secondly,the attention layer was used to allocate the corresponding weight of the extracted deep information from the domain names.Finally,the different weights of features in domain names were put into the output layer to complete the tasks of detection and classification.Our extensive experimental results demonstrate the effectiveness of the proposed model,both on regular DGA domains and DGA that hard to detect such as wordlist-based and part-wordlist-based ones.To be precise,we got a F1 score of 98.79% for the detection and macro average precision and recall of 83% for the classification task of DGA domain names.