Antenna selection for MIMO system based on pattern recognition

This paper proposes a novel Multiple-Input Multiple-Output (MIMO) transmission scheme based on Pattern Recognition (PR), which is termed as the PR aided Transmission Antenna Selection MIMO (PR-TAS aided MIMO). As the conventional TAS algorithms need to search all possible legitimate antenna subsets, they may impose some redundant calculations. In order to avoid this problem, we employ some pattern recognition methods to carry out the TAS algorithm in this paper. To be specific, two PR algorithms, namely the K-Nearest Neighbor (KNN) algorithm and the Support Vector Machine (SVM) algorithm, are introduced and redesigned to obtain a TAS with lower complexity but higher efficiency. Moreover, in order to improve the performance of the SVM, we propose a new feature extraction of channel matrix for the TAS. Our simulation results show that the proposed KNN and SVM based PR-TAS algorithms are capable of striking a flexible tradeoff between the complexity and the Bit Error Rate (BER), and the new feature can effectively improve the BER performance compared with the conventional feature extraction method.     

Tibetan Multi-Dialect Speech and Dialect Identity Recognition

Tibetan language has very limited resource for conventional automatic speech recognition so far. It lacks of enough data, sub-word unit, lexicons and word inventories for some dialects. And speech content recognition and dialect classification have been treated as two independent tasks and modeled respectively in most prior works. But the two tasks are highly correlated. In this paper, we present a multi-task WaveNet model to perform simultaneous Tibetan multi-dialect speech recognition and dialect identification. It avoids processing the pronunciation dictionary and word segmentation for new dialects, while, in the meantime, allows training speech recognition and dialect identification in a single model. The experimental results show our method can simultaneously recognize speech content for different Tibetan dialects and identify the dialect with high accuracy using a unified model. The dialect information used in output for training can improve multi-dialect speech recognition accuracy, and the low-resource dialects got higher speech content recognition rate and dialect classification accuracy by multi-dialect and multi-task recognition model than task-specific models.     

Facial expression recognition of intercepted video sequences based on feature point movement trend and feature block texture variation

Facial Expression Recognition (FER) is an important subject of human–computer interaction and has long been a research area of great interest. Accurate Facial Expression Sequence Interception (FESI) and discriminative expression feature extraction are two enormous challenges for the video-based FER. This paper proposes a framework of FER for the intercepted video sequences by using feature point movement trend and feature block texture variation. Firstly, the feature points are marked by Active Appearance Model (AAM) and the most representative 24 of them are selected. Secondly, facial expression sequence is intercepted from the face video by determining two key frames whose emotional intensities are minimum and maximum, respectively. Thirdly, the trend curve which represents the Euclidean distance variations between any two selected feature points is fitted, and the slopes of specific points on the trend curve are calculated. Finally, combining Slope Set which is composed by the calculated slopes with the proposed Feature Block Texture Difference (FBTD) which refers to the texture variation of facial patch, the final expressional feature are formed and inputted to One-dimensional Convolution Neural Network (1DCNN) for FER. Five experiments are conducted in this research, and three average FER rates 95.2%, 96.5%, and 97% for Beihang University (BHU) facial expression database, MMI facial expression database, and the combination of two databases, respectively, have shown the significant advantages of the proposed method over the existing ones.     

基于张量分解融合RGB-D图像的物体识别

为了充分利用RGB-D图像的深度图像信息，提出了基于张量分解的物体识别方法。首先将RGB-D图像构造成一个四阶张量，然后将该四阶张量分解为一个核心张量和四个因子矩阵，再利用相应的因子矩阵将原张量进行投影，获得融合后的RGB-D数据，最后输入到卷积神经网络中进行识别。RGB-D数据集中三组相似物体的识别结果表明，利用张量分解融合RGB-D图像的物体识别准确率高于未采用张量分解的物体识别准确率，并且单一错分实例的准确率最高可提升99%。     

The Internet shutdown muzzle(s) freedom of expression in Ethiopia: competing narratives

Freedom of expression is a foundational stone for democracy and embraces a number of other rights, including – freedom to seek, receive, and impart information using any medium. Around 16 million Ethiopians, comprising 15.4% of the total population, use the Internet. The Ethiopian Internet shutdown saga has been marked by competing narratives. On one hand, the government has been using two governing narratives, i.e. national security narrative and ‘economic growth’ narrative to justify shutdowns of the Internet. But, the individuals’ self-expressions on social media find itself as competing narrative. Since 2016, the Internet has been closed more than six times under the cloak of national security, quelling rising protests, controlling strikes, and exam cheating. As such, freedom of expression of millions has been muzzled. The article argues blanket Internet shutdowns do not meet the caveats of human rights law.     

基于支持向量机的RSA电磁旁路分析方法

针对在旁路分析过程中由于噪声的影响，不能直接观察出RSA的加密过程这一问题，提出结合支持向量机的旁路分析方法，从分类的角度对RSA二进制模幂运算中的平方、乘法操作进行识别，根据密钥与操作的相关性，推断出RSA二进制密钥序列。基于此，对移动设备PCM-9589F凌动主板进行了RSA电磁旁路分析研究，使用串口通信技术实现了对目标设备CPU的旁路电磁信号的采集，并采用基于边界的单类支持向量机的方法，以分类的正确率为70%为阈值，实现了对RSA二进制模幂运算中平方、乘法操作的识别，提取了OpenSSL加密库中1?024位RSA加密算法的二进制密钥序列。相比于传统的简单分析方法，克服了因旁路信号质量低而无法破解密钥的难题。     

基于结构化遮挡编码和极限学习机的局部遮挡人脸识别

提出使用结构化遮挡编码（SOC）结合极限学习机（ELM）的算法来处理人脸识别中的遮挡问题。首先，使用SOC去除图像上的遮挡物，将遮挡物体与人脸分离开；同时，通过局部性约束字典（LCD）来估计遮挡物的位置，建立遮挡字典和人脸字典。然后，将建立好的人脸字典矩阵进行归一化处理，并利用ELM对归一化的数据进行分类识别。最后，在AR人脸库上进行的仿真实验结果表明，所提方法对不同遮挡物和不同区域遮挡的图像具有较好的识别率和鲁棒性。     

基于多尺度双线性卷积神经网络的多角度下车型精细识别

针对多角度下车辆出现一定的尺度变化和形变导致很难被准确识别的问题，提出基于多尺度双线性卷积神经网络（MS-B-CNN）的车型精细识别模型。首先，对双线性卷积神经网络（B-CNN）算法进行改进，提出MS-B-CNN算法对不同卷积层的特征进行了多尺度融合，以提高特征表达能力；此外，还采用基于中心损失函数与Softmax损失函数联合学习的策略，在Softmax损失函数基础上分别对训练集每个类别在特征空间维护一个类中心，在训练过程中新增加样本时，网络会约束样本的分类中心距离，以提高多角度情况下的车型识别的能力。实验结果显示，该车型识别模型在CompCars数据集上的正确率达到了93.63%，验证了模型在多角度情况下的准确性和鲁棒性。