微表情是一个人试图隐藏内心真实情感却又不由自主流露出的不易被察觉的面部表情。与一般面部表情相比,微表情最显著的特点是持续时间短、强度弱,往往难以有效识别。文中提出了一种基于LBP-TOP(Local Binary Pattern from Three Orthogo...微表情是一个人试图隐藏内心真实情感却又不由自主流露出的不易被察觉的面部表情。与一般面部表情相比,微表情最显著的特点是持续时间短、强度弱,往往难以有效识别。文中提出了一种基于LBP-TOP(Local Binary Pattern from Three Orthogonal Planes)特征和支持向量机(Support Vector Machine,SVM)分类器的微表情识别方法。首先,采用LBP-TOP算子来提取微表情特征;然后,提出一种基于ReliefF与局部线性嵌入(Locally Linear Embedding,LLE)流形学习算法相结合的特征选择算法,对提取的LBP-TOP特征向量进行降维;最后,使用径向基函数(Radial Basis Function,RBF)核的SVM分类器进行分类,将测试样本图像序列的微表情分为5类:高兴、厌恶、压抑、惊讶、其他。在CASME Ⅱ微表情数据库上采用"留一人交叉验证"(Leave-One-Subject-Out Cross Validation,LOSO-CV)的方式进行了实验,可得到58.98%的分类准确率。实验结果表明了该算法的有效性。展开更多
微表情区别于普通的面部表情,具有持续时间短、面部强度低的特点,往往难以有效识别,制约了该领域的研究。针对上述难点,提出一种新颖的特征结合方法。采用全局光流技术在相邻帧间进行计算,得到微弱光流,通过传递前后各帧的运动信息,在...微表情区别于普通的面部表情,具有持续时间短、面部强度低的特点,往往难以有效识别,制约了该领域的研究。针对上述难点,提出一种新颖的特征结合方法。采用全局光流技术在相邻帧间进行计算,得到微弱光流,通过传递前后各帧的运动信息,在相隔多帧的两幅图像间体现更为明显的变化,解决了短历时和动作微弱的难题;将光流特征与LBP-TOP(Local Binary Patterns from Three Orthogonal Planes)算子提取的时空局部纹理特征相结合,补充描述人脸大多数区域的细节信息。选择随机森林分类器进行实验,实验结果表明,两种特征具有很好的互补性,在CASMEII数据库下,能识别5类情感,准确率由40.50%提高至64.46%,类间区分度也有相应改善。展开更多
提出基于分块LBP-TOP(Local Binary Patterns from Three Orthogonal Planes)特征和改进的加权稀疏表示分类解决微表情识别与专用线车辆状态检测问题。首先利用LBP-TOP特征描述符对从分块图像中选择出的有效块进行提取特征,将提取的特...提出基于分块LBP-TOP(Local Binary Patterns from Three Orthogonal Planes)特征和改进的加权稀疏表示分类解决微表情识别与专用线车辆状态检测问题。首先利用LBP-TOP特征描述符对从分块图像中选择出的有效块进行提取特征,将提取的特征作为字典,采取加权稀疏表示(Weighted Sparse Representation,WSRC)和对偶增广拉格朗日乘子法(Dual Augmented Lagrange Multiplier,DALM)相结合的算法(WSRC_DALM)进行稀疏表示分类;然后利用不同尺寸的块划分图像,选择有效块提取特征,特征融合后参与分类。在CASMEⅡ与SAMM表情数据库上采用"留一人交叉验证"(Leave One Subject Out, LOSO)的分类方法进行5分类,得到的识别率分别达到了77.30%与58.82%,在车辆状态检测检测数据库上的实验达到了84.60%的检测率。实验结果表明了所提出算法的有效性。展开更多
Background:The accurate identification of cardiac abnormalities is essential for proper diagnosis and effective treatment of cardiovascular diseases.Method:This work introduces an advanced methodology for detecting ca...Background:The accurate identification of cardiac abnormalities is essential for proper diagnosis and effective treatment of cardiovascular diseases.Method:This work introduces an advanced methodology for detecting cardiac abnormalities and estimating electrocardiographic age(ECG Age)using sophisticated signal processing and deep learning techniques.This study looks at six main heart conditions found in 12-lead electrocardiogram(ECG)data.It addresses important issues like class imbalances,missing lead scenarios,and model generalizations.A modified residual neural network(ResNet)architecture was developed to enhance the detection of cardiac abnormalities.Results:The proposed ResNet demonst rated superior performance when compared with two linear models and an alternative ResNet architectures,achieving an overall classification accuracy of 91.25%and an F1 score of 93.9%,surpassing baseline models.A comprehensive lead loss analysis was conducted,evaluating model performance across 4096 combinations of missing leads.The results revealed that pulse rate-based factors remained robust with up to 75%lead loss,while block-based factors experienced significant performance declines beyond the loss of four leads.Conclusion:This analysis highlighted the importance of addressing lead loss impacts to maintain a robust model.To optimize performance,targeted training approaches were developed for different conditions.Based on these insights,a grouping strategy was implemented to train specialized models for pulse rate-based and block-based conditions.This approach resulted in notable improvements,achieving an overall classification accuracy of 95.12%and an F1 score of 95.79%.展开更多
为自动识别视频中表情类别,提出基于面部块表情特征编码的视频表情识别方法框架。检测并精确定位视频中人脸关键点位置,以检测到的关键点为中心,提取面部显著特征块。沿着时间轴方向,对面部各特征块提取LBP-TOP(local binary pattern fr...为自动识别视频中表情类别,提出基于面部块表情特征编码的视频表情识别方法框架。检测并精确定位视频中人脸关键点位置,以检测到的关键点为中心,提取面部显著特征块。沿着时间轴方向,对面部各特征块提取LBP-TOP(local binary pattern from three orthogonal planes)动态特征描述子,将这些描述子作为表情特征并输入Adaboost分类器进行训练和识别,预测视频表情类型。在国际通用表情数据库BU-4DFE的纹理图像上进行测试,取得了81.2%的平均识别率,验证了所提算法的有效性,与同领域其它主流算法相比,其具有很强的竞争性。展开更多
MSU Stego Video is a public video steganographic tool, which has strong robustness and is regarded as a real video steganographic tool. In order to increase the detection rate, this paper proposes a new steganoalysis ...MSU Stego Video is a public video steganographic tool, which has strong robustness and is regarded as a real video steganographic tool. In order to increase the detection rate, this paper proposes a new steganoalysis method against MSU, which uses the chessboard character of MSU embedded video, proposes a down-sample block-based collusion method to estimate the original frame and checks the chessboard mode of the different frame between tested frame and estimated frame to detect MSU steganographic evidences. To reduce the error introduced by severe movement of the video content, a method that abandons severe motion blocks from detecting is proposed. The experiment results show that the false negative rate of the proposed algorithm is lower than 5%, and the false positive rate is lower than 2%. Our algorithm has significantly better performance than existing algorithms. Especially to the video that has fast motion, the algorithm has more remarkable performance.展开更多
By combining the AdaBoost modular locality preserving projection （AMLPP） algorithm and the locally linear regression （LLR） algorithm, a novel pose-invariant algorithm is proposed to realize high-accuracy face reco...By combining the AdaBoost modular locality preserving projection （AMLPP） algorithm and the locally linear regression （LLR） algorithm, a novel pose-invariant algorithm is proposed to realize high-accuracy face recognition under different poses. In the training stage of this algorithm, the AMLPP is employed to select the crucial frontal blocks and construct effective strong classifier. According to the selected frontal blocks and the corresponding non-frontal blocks, LLR is then applied to learn the linear mappings which will be used to convert the non-frontal blocks to visual frontal blocks. During the testing of the learned linear mappings, when a non-frontal face image is inputted, the non-frontal blocks corresponding to the selected frontal blocks are extracted and converted to the visual frontal blocks. The generated virtual frontal blocks are finally fed into the strong classifier constructed by AMLPP to realize accurate and efficient face recognition. Our algorithm is experimentally compared with other pose-invariant face recognition algorithms based on the Bosphorus database. The results show a significant improvement with our proposed algorithm.展开更多
Through the analysis of the international definition and classification of slums,this paper explores the development of China's social housing system and the renovation of the Old City of Kashgar.It argues that on...Through the analysis of the international definition and classification of slums,this paper explores the development of China's social housing system and the renovation of the Old City of Kashgar.It argues that one of the issues in China's social housing system is to solve the problems of the scarcity of spatial elements and the lack of developmental driving force in large mixed communities of the Han and minority nationalities.Then it examines the elements of renovation and approaches based on a survey of the local residents in different parts of Kashgar City.Comparing the international development of traditional residential quarters and block-based communities,the paper points out that the block-based community is preferred for its impartiality and sustainability,and applies this mode to the renovation of the Old City of Kashgar in the form of design guidelines.展开更多
Due to coarse quantization, block-based discrete cosine transform(BDCT) compression methods usually suffer from visible blocking artifacts at the block boundaries. A novel efficient de-blocking method in DCT domain is...Due to coarse quantization, block-based discrete cosine transform(BDCT) compression methods usually suffer from visible blocking artifacts at the block boundaries. A novel efficient de-blocking method in DCT domain is proposed. A specific criterion for edge detection is given, one-dimensional DCT is applied on each row of the adjacent blocks and the shifted block in smooth region, and the transform coefficients of the shifted block are modified by weighting the average of three coefficients of the block. Mean square difference of slope criterion is used to judge the efficiency of the proposed algorithm. Simulation results show that the new method not only obtains satisfactory image quality, but also maintains high frequency information.展开更多
This paper consists of a lossy image compression algorithm dedicated to the medical images doing comparison of RGB and YCbCr color space. Several lossy/lossless transform coding techniques are used for medical image c...This paper consists of a lossy image compression algorithm dedicated to the medical images doing comparison of RGB and YCbCr color space. Several lossy/lossless transform coding techniques are used for medical image compression. Discrete Wavelet Transform (DWT) is one such widely used technique. After a preprocessing step (remove the mean and RGB to YCbCr transformation), the DWT is applied and followed by the bisection method including thresholding, the quantization, dequantization, the Inverse Discrete Wavelet Transform (IDWT), YCbCr to RGB transform of mean recovering. To obtain the best compression ratio (CR), the next step encoding algorithm is used for compressing the input medical image into three matrices and forward to DWT block a corresponding containing the maximum possible of run of zeros at its end. The last step decoding algorithm is used to decompress the image using IDWT that is applied to get three matrices of medical image.展开更多
Tracking objects that undergo abrupt appearance changes and heavy occlusions is a challenging problem which conventional tracking methods can barely handle. To address the problem, we propose an online structure learn...Tracking objects that undergo abrupt appearance changes and heavy occlusions is a challenging problem which conventional tracking methods can barely handle. To address the problem, we propose an online structure learning algorithm that contains three layers： an object is represented by a mixture of online structure models （OSMs） which are learnt from block-based online random forest classifiers （BORFs）. BORFs are able ~o handle occlusion problems since they model local appearances of the target. To further improve the tracking accuracy and reliability, the algorithm utilizes mixture relational models （MRMs） as multi-mode context information to integrate BORFs into OSMs. Furthermore, the mixture construction of OSMs can avoid over-fitting effectively and is more flexible to describe targets. Fusing BORFs with MRMs, OSMs capture the discriminative parts of the target, which guarantees the reliability and robustness of our tracker. In addition, OSMs incorporate with block occlusion reasoning to update our BORFs and MRMs, which can deal with appearance changes and drifting problems effectively. Experiments on challenging videos show that the proposed tracker performs better than several state-of-the-art algorithms.展开更多
文摘微表情区别于普通的面部表情,具有持续时间短、面部强度低的特点,往往难以有效识别,制约了该领域的研究。针对上述难点,提出一种新颖的特征结合方法。采用全局光流技术在相邻帧间进行计算,得到微弱光流,通过传递前后各帧的运动信息,在相隔多帧的两幅图像间体现更为明显的变化,解决了短历时和动作微弱的难题;将光流特征与LBP-TOP(Local Binary Patterns from Three Orthogonal Planes)算子提取的时空局部纹理特征相结合,补充描述人脸大多数区域的细节信息。选择随机森林分类器进行实验,实验结果表明,两种特征具有很好的互补性,在CASMEII数据库下,能识别5类情感,准确率由40.50%提高至64.46%,类间区分度也有相应改善。
文摘Background:The accurate identification of cardiac abnormalities is essential for proper diagnosis and effective treatment of cardiovascular diseases.Method:This work introduces an advanced methodology for detecting cardiac abnormalities and estimating electrocardiographic age(ECG Age)using sophisticated signal processing and deep learning techniques.This study looks at six main heart conditions found in 12-lead electrocardiogram(ECG)data.It addresses important issues like class imbalances,missing lead scenarios,and model generalizations.A modified residual neural network(ResNet)architecture was developed to enhance the detection of cardiac abnormalities.Results:The proposed ResNet demonst rated superior performance when compared with two linear models and an alternative ResNet architectures,achieving an overall classification accuracy of 91.25%and an F1 score of 93.9%,surpassing baseline models.A comprehensive lead loss analysis was conducted,evaluating model performance across 4096 combinations of missing leads.The results revealed that pulse rate-based factors remained robust with up to 75%lead loss,while block-based factors experienced significant performance declines beyond the loss of four leads.Conclusion:This analysis highlighted the importance of addressing lead loss impacts to maintain a robust model.To optimize performance,targeted training approaches were developed for different conditions.Based on these insights,a grouping strategy was implemented to train specialized models for pulse rate-based and block-based conditions.This approach resulted in notable improvements,achieving an overall classification accuracy of 95.12%and an F1 score of 95.79%.
文摘为自动识别视频中表情类别,提出基于面部块表情特征编码的视频表情识别方法框架。检测并精确定位视频中人脸关键点位置,以检测到的关键点为中心,提取面部显著特征块。沿着时间轴方向,对面部各特征块提取LBP-TOP(local binary pattern from three orthogonal planes)动态特征描述子,将这些描述子作为表情特征并输入Adaboost分类器进行训练和识别,预测视频表情类型。在国际通用表情数据库BU-4DFE的纹理图像上进行测试,取得了81.2%的平均识别率,验证了所提算法的有效性,与同领域其它主流算法相比,其具有很强的竞争性。
基金Supported by the National Natural Science Foundation of China(60970114)Doctoral Fund of Ministry of Education of China(20110141130006)
文摘MSU Stego Video is a public video steganographic tool, which has strong robustness and is regarded as a real video steganographic tool. In order to increase the detection rate, this paper proposes a new steganoalysis method against MSU, which uses the chessboard character of MSU embedded video, proposes a down-sample block-based collusion method to estimate the original frame and checks the chessboard mode of the different frame between tested frame and estimated frame to detect MSU steganographic evidences. To reduce the error introduced by severe movement of the video content, a method that abandons severe motion blocks from detecting is proposed. The experiment results show that the false negative rate of the proposed algorithm is lower than 5%, and the false positive rate is lower than 2%. Our algorithm has significantly better performance than existing algorithms. Especially to the video that has fast motion, the algorithm has more remarkable performance.
基金Supported by the National Natural Science Foundation of China(60772066)
文摘By combining the AdaBoost modular locality preserving projection （AMLPP） algorithm and the locally linear regression （LLR） algorithm, a novel pose-invariant algorithm is proposed to realize high-accuracy face recognition under different poses. In the training stage of this algorithm, the AMLPP is employed to select the crucial frontal blocks and construct effective strong classifier. According to the selected frontal blocks and the corresponding non-frontal blocks, LLR is then applied to learn the linear mappings which will be used to convert the non-frontal blocks to visual frontal blocks. During the testing of the learned linear mappings, when a non-frontal face image is inputted, the non-frontal blocks corresponding to the selected frontal blocks are extracted and converted to the visual frontal blocks. The generated virtual frontal blocks are finally fed into the strong classifier constructed by AMLPP to realize accurate and efficient face recognition. Our algorithm is experimentally compared with other pose-invariant face recognition algorithms based on the Bosphorus database. The results show a significant improvement with our proposed algorithm.
文摘Through the analysis of the international definition and classification of slums,this paper explores the development of China's social housing system and the renovation of the Old City of Kashgar.It argues that one of the issues in China's social housing system is to solve the problems of the scarcity of spatial elements and the lack of developmental driving force in large mixed communities of the Han and minority nationalities.Then it examines the elements of renovation and approaches based on a survey of the local residents in different parts of Kashgar City.Comparing the international development of traditional residential quarters and block-based communities,the paper points out that the block-based community is preferred for its impartiality and sustainability,and applies this mode to the renovation of the Old City of Kashgar in the form of design guidelines.
基金Science and Technology Project of Guangdong Province(2006A10201003) 2005 Jinan University StartupProject(51205067) Soft Science Project of Guangdong Province(2006B70103011)
文摘Due to coarse quantization, block-based discrete cosine transform(BDCT) compression methods usually suffer from visible blocking artifacts at the block boundaries. A novel efficient de-blocking method in DCT domain is proposed. A specific criterion for edge detection is given, one-dimensional DCT is applied on each row of the adjacent blocks and the shifted block in smooth region, and the transform coefficients of the shifted block are modified by weighting the average of three coefficients of the block. Mean square difference of slope criterion is used to judge the efficiency of the proposed algorithm. Simulation results show that the new method not only obtains satisfactory image quality, but also maintains high frequency information.
文摘This paper consists of a lossy image compression algorithm dedicated to the medical images doing comparison of RGB and YCbCr color space. Several lossy/lossless transform coding techniques are used for medical image compression. Discrete Wavelet Transform (DWT) is one such widely used technique. After a preprocessing step (remove the mean and RGB to YCbCr transformation), the DWT is applied and followed by the bisection method including thresholding, the quantization, dequantization, the Inverse Discrete Wavelet Transform (IDWT), YCbCr to RGB transform of mean recovering. To obtain the best compression ratio (CR), the next step encoding algorithm is used for compressing the input medical image into three matrices and forward to DWT block a corresponding containing the maximum possible of run of zeros at its end. The last step decoding algorithm is used to decompress the image using IDWT that is applied to get three matrices of medical image.
基金supported in part by the National Natural Science Foundation of China under Grant No. 61075026the National Basic Research 973 Program of China under Grant No. 2011CB302203.
文摘Tracking objects that undergo abrupt appearance changes and heavy occlusions is a challenging problem which conventional tracking methods can barely handle. To address the problem, we propose an online structure learning algorithm that contains three layers： an object is represented by a mixture of online structure models （OSMs） which are learnt from block-based online random forest classifiers （BORFs）. BORFs are able ~o handle occlusion problems since they model local appearances of the target. To further improve the tracking accuracy and reliability, the algorithm utilizes mixture relational models （MRMs） as multi-mode context information to integrate BORFs into OSMs. Furthermore, the mixture construction of OSMs can avoid over-fitting effectively and is more flexible to describe targets. Fusing BORFs with MRMs, OSMs capture the discriminative parts of the target, which guarantees the reliability and robustness of our tracker. In addition, OSMs incorporate with block occlusion reasoning to update our BORFs and MRMs, which can deal with appearance changes and drifting problems effectively. Experiments on challenging videos show that the proposed tracker performs better than several state-of-the-art algorithms.
