Search

In this letter, it is proposed the hyperspectral image classification method based on the convolutional neural network, which is trained jointly by the reconstruction and discriminative loss functions. In the network, small convolutional kernels are cascaded with the pooling operator to perform feature abstraction, and a decoding channel composed of the deconvolutional and unpooling operators is established. The unsupervised reconstruction, performed by the decoding channel, not only introduces priors to the network training but also is made use to enhance the discriminability of the abstracted features by the control gate. By the experiments, it is shown that the proposed method performs better than the state-of-the-art neural network-based classification methods.

The current situation, hot spot and trend of coronavirus research in public health of China are discussed, which provides reference for the future research and direction of coronavirus. Based on statistical analysis methods such as bibliometric and knowledge mapping, the visual software CiteSpace was used to analyze the knowledge mapping constructed by keywords, clustering, authors and research institutions. Domestic scholars have analyzed and discussed virus tracing, transmission mechanism, virus detection, disease diagnosis and treatment, drug research and development, epidemic situation assessment and other aspects of coronavirus. It has become a trend to use new technologies to improve research and development programs. In addition, traditional Chinese medicine will play an increasingly important role in the future research of coronavirus in public health.

It is important and necessary to obtain high-frequency information and texture details in the image reconstruction applications, such as image super-resolution. Hence, it is proposed the multi-scale fusion network (MCFN) in this paper. In the network, three pathways are designed for different receptive fields and scales, which are expected to obtain more texture details. Meanwhile, the local and global residual learning strategies are employed to prevent overfitting and to improve reconstruction quality. Compared with the classic convolutional neural network-based algorithms, the proposed method achieves better numerical and visual effects.

In order to address the impact of large-angle posture changes on face recognition performance, we propose a contour symmetric constraint-generative adversarial network (CSC-GAN) for the multi-pose face recognition. The method employs the convolutional network as the generator for face pose recovery, which introduces the global information of the constrained pose recovery of positive face contour histogram. Meanwhile, the original positive face is used as the discriminator, and the symmetric loss function is added to optimize the learning ability of the network. The positive face with gesture recovery is obtained by striking the balance between training of the generator and discriminator. Then we employed the nearest neighbor classifier to identify. The experimental results show that CSC-GAN obtained good posture reconstruction texture information on the multi-pose face reconstruction. Compared with the traditional deep learning method and 3D method, it also achieves higher recognition rate.

Suggestions on Chinese literature cataloging of Mainland, Taiwan, Hong Kong and Macao are put forward, with a view to achieving world-wide sharing of Chinese bibliographic metadata. Cataloging rules and MARC format of Chinese literatures are compared in four places of China. The respective characteristics and the present situation are analyzed. According to the analysis results, CNMRAC, CMARC and MARC21 have their own advantages and disadvantages. A unified resource description framework is needed for the processing of Chinese literature cataloging. BIBFRAME is a more suitable unified resource description framework which needs further research and efforts to adapt to Chinese literatures in digital library.

To improve the accuracy in low resolution face recognition, a method based on super-resolution joint feature mapping is proposed. Firstly, a two-branch convolutional neural network is designed to extract features of high and low resolution face images. A super-resolution enhanced network cascading feature extraction network is used for feature mapping of low resolution face images. In this way, the high frequency information of low resolution image can be reconstructed, and features are extracted. Secondly, a fusion loss method is utilized, in which the loss of cosine and the image reconstruction are weighted and fusioned to increase the cosine similarity between image features of different resolutions. Finally, the experimental results based on FERET dataset validate that the test accuracy of two-branch framework is up to 98.2%, 99.1%, 99.5% with resolutions of 20× 20, 24× 24, and 36× 36 obtained by smooth downsampling. The proposed model outperforms up-to-date low resolution face recognition methods.

This paper proposes an improved method for feature selection based on traditional mutual information by establishing domain feature words which utilize the differences in the representation of a word in different classes. By the method, we can reselect the feature set out of the established one based on the traditional mutual information. It not only reduces the dimension of the vector but also represents the text more effectively. At the same time, a text similarity calculation system is designed in this paper. Finally, the experimental results show that the improved feature extraction method is superior to the traditional mutual information and the system has a good performance.

Resilient Packet Ring (RPR) is the metropolitan area network (MAN) architecture of based on optimization of packet service, its ring fabric is the part of register insert ring. Register insert ring has the characteristics of time-division multiplexing and space-division multiplexing, and the schedule of ring fabric is simple. It is suitable for fabric of switch. The article uses register insert ring for ring fabric, and models the structure of ring switching according to queuing theory and analyses its performance. The results show that the structure of register insertion ring is suitable for high- speed packet switching.

Automatic detection of an unusual event in video sequence has an interesting application in security surveillance. This paper proposed a method to detect a gathering event without tracking individuals by using background subtracted blobs as the event feature and dividing the video frame into blocks. The feature sequences are encoded with hidden markov model to detect the gathering event. The experimental results show the effectiveness of the proposed approach.

To resist various attacks, a Boolean function used in symmetric cipher systems should satisfy multiple cryptographic criteria. In this paper, a new generalized construction method for Boolean functions which satisfy multiple cryptographic criteria is provided. These multiple cryptographic criteria include balanced, high nonlinearity, high algebraic degree, propagation criterion of order l (PC(l)) and nonexistence of nonzero linear structures. The construction is based on the use of linear error-correcting code. Given a linear code and its dual code, we show that it is possible to obtain these cryptographic functions.

Dempster-Shafer (D-S) evidence theory is an effective method used for Distributed Decision Fusion (DDF). However, constructing a basic probability assignment function (BPAF) is a challenging task for practical fusion applications. This paper proposes a novel method for DDF of target identification. In the presented algorithm, Support Vector Machine (SVM) is utilized as the identifier, and the basic probability assignment function is constructed accordingly. The final fused results are obtained by applying the D-S evidence theory to the outputs of the identifiers. The experimental results show that the proposed method provides higher accuracy and reliability when it is compared with existing voting schemes.

Computer electrocardiography (EGG) is a fundamental diagnostic method for both contour and rhythm analysis. For patients with Hypertrophic Cardiomyopathy (HCM), there are more or less abnormal ECG findings. However, to diagnose HCM through ECG is a difficult task even for experienced cardiologist. Backpropagation has been used for ECG classification with the number of hidden units chosen heuristically. In this paper, the hidden unit number is selected by a new criteria obtained from the so-call Bayesian Ying-Yang learning theory and applied in ECG classification to diagnose HCM. Experiments have shown that the selected number is highly consistent with the minimal generalization error and the corresponding architecture show best classification performance.