Convolutional neural networks are a hot research topic in finger vein recognition. However, most existing research focuses on increasing the depth and width of the convolutional neural network to improve the network's performance, which has a specific requirement on the computer's computational power. To reduce the computational burden, a lightweight convolutional attention model (LCAModel) is proposed for finger vein recognition to achieve more accurate visual structure capture by exploiting the sensitivity of the attention mechanism to features. First, the attentional model (AModel) is proposed to extract representative features of images, which mainly utilizes the adaptive mapping on space and channels of the convolutional block attention module (CBAM) to make features distinguishable. In addition, considering the integrity of the features, a convolutional model (CModel) is designed to supplement the features of AModel. Finally, the features obtained from AModel and CModel are fused using an adaptive weighting mechanism to make the features more complete. Here, the obtained features are provided into a support vector machine (SVM) for image classification. The experiments are carried out on two publicly available databases, demonstrating that the proposed network structure requires less computing power and performs better.

Abstract With the development of the information era, the study of biometric keys has attracted increasing attention. However, the performance and safety of the proposed methods seem challenging to meet the needs of real applications. To solve this problem, this paper presents a finger vein biometric key system. The proposed system proposes an improved network in the feature extraction stage, which combines the coordinate attention mechanism and SkipNet to obtain richer feature information. An improved, reliable feature selection method is introduced to acquire more discriminative finger vein features. A Lagrange interpolation method then binds the user's finger vein features and keys. Finally, experiments are carried out on the public USM, Poly, and Net64 databases. The results indicate that we acquire a 1024-bit key with 0.049% FAR and 0% FRR in Net64 and 0% FAR and FRR in USM and Poly. Additionally, the system's security is analyzed from key randomness, system information leak attack, brute force attack, cross-matching attack, and spoofing attack. Experimental and theoretical analyses show that the proposed system has good accuracy and security.

Convolutional neural networks have become a popular research in the field of finger vein recognition because of their powerful image feature representation. However, most researchers focus on improving the performance of the network by increasing the CNN depth and width, which often requires high computational effort. Moreover, we can notice that not only the importance of pixels in different channels is different, but also the importance of pixels in different positions of the same channel is different. To reduce the computational effort and to take into account the different importance of pixels, we propose a lightweight convolutional neural network with a convolutional block attention module (CBAM) for finger vein recognition, which can achieve a more accurate capture of visual structures through an attention mechanism. First, image sequences are fed into a lightweight convolutional neural network we designed to improve visual features. Afterwards, it learns to assign feature weights in an adaptive manner with the help of a convolutional block attention module. The experiments are carried out on two publicly available databases and the results demonstrate that the proposed method achieves a stable, highly accurate, and robust performance in multimodal finger recognition.

The texture of finger veins is distributed in a network structure, which can be described as regional texture feature. In the image preprocessing stage, noise generated by segmentation algorithm will lead to the loss of texture structure information. Local binary pattern (LBP) feature extraction, which does not require segmentation of images, can effectively reveal local texture features and is robust to monotonic changes in grayscale. However, the LBP operator has two obvious shortcomings: (1) the microscopic limitation: it is easy to lose local information; (2) the feature unity: it will lead to the loss of other feature information. To tack these problems, this paper proposes a multi-feature partitioned local binary pattern (MFPLBP) operator for finger vein recognition. The concept of multi-feature partition is employed to extend the traditional LBP operator. Through the partition processing of the finger vein feature image, the global and local grasp of the image is enhanced, and the influence of local noise on the overall recognition accuracy is weakened. Additionally, the idea of multi-feature fusion is used to make up for the singleness of traditional algorithms. In image recognition, the histogram cross-check is used to judge the similarity of the vein feature histogram. Finally, the experiment showed that the recognition rate of this method has increased by about 13% compared with LBP, and it has increased by about 2% compared with partitioned local binary pattern (PLBP) and traditional multi-scale LBP.

Finger vein recognition has drawn increasing attention as one of the most popular and promising biometrics due to its high distinguishing ability, security, and non-invasive procedure. The main idea of traditional schemes is to directly extract features from finger vein images and then compare features to find the best match. However, the features extracted from images contain much redundant data, while the features extracted from patterns are greatly influenced by image segmentation methods. To tackle these problems, this paper proposes a new finger vein recognition algorithm by generating code. The proposed method does not require an image segmentation algorithm, is simple to calculate, and has a small amount of data. Firstly, the finger vein images were divided into blocks to calculate the mean value. Then, the centrosymmetric coding was performed using the matrix generated by blocking and averaging. The obtained codewords were concatenated as the feature codewords of the image. The similarity between vein codes is measured by the ratio of minimum Hamming distance to codeword length. Extensive experiments on two public finger vein databases verify the effectiveness of the proposed method. The results indicate that our method outperforms the state-of-the-art methods and has competitive potential in performing the matching task.

Finger vein recognition has drawn increasing attention as one of the most popular and promising biometrics due to its high distinguishes ability, security and non-invasive procedure. The main idea of traditional schemes is to directly extract features from finger vein images or patterns and then compare features to find the best match. However, the features extracted from images contain much redundant data, while the features extracted from patterns are greatly influenced by image segmentation methods. To tack these problems, this paper proposes a new finger vein recognition by generating code. The proposed method does not require an image segmentation algorithm, is simple to calculate and has a small amount of data. Firstly, the finger vein images were divided into blocks to calculate the mean value. Then the centrosymmetric coding is performed by using the generated eigenmatrix. The obtained codewords are concatenated as the feature codewords of the image. The similarity between vein codes is measured by the ratio of minimum Hamming distance to codeword length. Extensive experiments on two public finger vein databases verify the effectiveness of the proposed method. The results indicate that our method outperforms the state-of-theart methods and has competitive potential in performing the matching task.

Aiming at the problems of Chinese word segmentation (CWS) and part-of-speech (POS) tagging of quantitative units in elementary mathematics application problems, an improved CWS algorithm is proposed. By compiling the revised dictionary about the quantitative units, designing a segmentation algorithm and a merging algorithm to segment and then merge the matched target words, the problems of CWS and POS tagging of quantitative units are corrected. The experimental results show that the improved CWS algorithm can improve the accuracy of CWS and POS tagging of quantitative units to 98%.

以比特币为代表的加密数字货币涉及的所有交易会存储在一个被称为区块链的公共数据库中.随着加密数字货币的普及应用,货币丢失、被盗等安全性问题愈发突出.出于便利性和安全性考虑,加密数字货币通常通过钱包来进行使用.钱包中存储的最重要的数据是用户的私钥,这也是加密数字货币所有权的唯一标识,一旦丢失或者被盗将带来无法挽回的损失.针对区块链钱包方案进行研究,根据私钥的储存和使用方式将钱包分为硬件钱包、软件钱包、托管钱包和门限钱包四种类型,并介绍了目前较流行的几种钱包方案.通过对各种钱包方案使用便利性、安全性、效率等方面的分析和比较,总结了各类钱包方案的优点以及存在的不足,并给出了未来钱包的发展趋势.

Finger vein recognition is a new generation biometric recognition technology and has wide potential applications.In order to locate the finger vein region for recognition,one important work is to detect the edges of the finger to segment the finger shape.Because the existing finger edge detect algorithms did not work well for low quality finger vein images,this paper proposed a novel algorithm.It consisted of three main steps:firstly,the algorithm searched an inner line of the finger and it divided the finger vein image by upper part and lower part.Then,the algorithm detected the upper edge and lower edge of the finger by utilizing two horizontal edge detection templates in upper part and lower part respectively.Finally,the algorithm detected the wrong sections of the edges and fixed them by a series of technologies such as interpolation,fit,and so on.It performed a series of finger vein images in this experiments.The results show that the proposed algorithm is accurate and robust.

Automatic finger vein recognition systems have attracted more attentions in recent years. In order to implement a high performance system, an important step is to localize the region of interest accurately. A problem in previous ROI localization methods is that some useful finger vein information is lost in the final cropped ROI region. In order to resolve this problem, a novel ROI extraction method for finger vein images is proposed in this paper. Finger edges are detected and adjusted to the horizontal direction, after that a modified sliding window is used in order to detect the distal inter-joint line of the finger. On the basis of the edges and the distal inter-phalangeal joint line of the finger, different from previous methods, an outer rectangle is used to crop the finger area to avoid the useful information loss. Based on our experimental dataset with 3132 finger vein images, the mean information loss rate for previous methods is 15.1% and there is no loss of information for our method. In order to evaluate the accuracy of our ROI extraction method, the similarity rate of intra-class is calculated, which is defined by the ratio of overlap area and the whole ROI area. And a mean similarity rate 96.3% is obtained in our experiments. Theoretical analysis and experimental results show that the proposed method is effective and accurate, and it is potentially beneficial for improving the performance of finger vein recognition system.

Serial analysis of gene expression (SAGE) is an efficient technique to produce a snapshot of the messenger RNA population in a sample. Clustering method has been widely used for SAGE data mining. Clustering SAGE data into different pattern groups can help to find potentially unknown functional gene groups in SAGE dataset. By incorporating a new published measurement (maximal information coefficient, MIC) into hierarchical clustering techniques, we present a clustering method named MicClustSAGE. The MIC can measure the pair-wise correlation coefficients between SAGE libraries. The presented method significant improvements the ability of clustering method in detecting specially tissue pattern of SAGE. In addition, we compared the results obtained by our method and hierarchical clustering with Pearson correlation. The experimental results exhibit the performance of the proposed method on several real-life SAGE datasets.

Research on the stochastic behavior of traffic flow is important to understand the intrinsic evolution rule of traffic system. On the basis of cellular automata model and traffic flow model with look-ahead potential, in this paper, a novel traffic flow model with weighted look-ahead potential is presented. By introducing the weighting coefficient into the look-ahead potential and endowing the potential of vehicle closer to itself with a greater weight, the modeling process is more suitable for the driver's random decision-making process which is based on the vehicle and enviroment situation in front of him in actual traffic. Complex high-density traffic behavior is reproduced by numerical simulations. The simulation results show that the weighting coefficient has an obvious effect on high-density traffic flux, and the weighted model is more conducive to keeping high traffic flux and road capacity while maintaining a high traffic density.

To discover relationships and associations rapidly in large-scale datasets, we propose a cross-platform tool for the rapid computation of the maximal information coefficient based on parallel computing methods. Through parallel processing, the provided tool can effectively analyze large-scale biological datasets with a markedly reduced computing time. The experimental results show that the proposed tool is notably fast, and is able to perform an all-pairs analysis of a large biological dataset using a normal computer. The source code and guidelines can be downloaded from https://github.com/HelloWorldCN/RapidMic .

How to apply the entropy in biometrics into the encryption and remote authentication schemes to simplify the management of keys is a hot research area. Utilizing Dodis's fuzzy extractor method and Liu's original signcryption scheme, a biometric identity based signcryption scheme is proposed in this paper. The proposed scheme is more efficient than most of the previous proposed biometric signcryption schemes for that it does not need bilinear pairing computation and modular exponentiation computation which is time consuming largely. The analysis results show that under the CDH and DL hard problem assumption, the proposed scheme has the features of confidentiality and unforgeability simultaneously.

Ring signature is one of most important type of digital signature. Usually the ring signature is generated by one signer, however sometimes we need the signature be generated by one subset from an access structure of a group. Moreover, we need to overcome the management problem of private keys in traditional signature schemes. In this paper, a novel biometric ring signature scheme for authorized subsets is proposed. The members of the authorized subsets can cooperatively sign the message based on their biometric identity and the verifier can verify if the ring signature is generated by one authorized subset in the access structure, however he cannot find out which authorized subset has really issued the signature. The analysis results show that the scheme is secure and easy to implement. (C) 2011 Published by Elsevier Ltd.

The signcryption scheme plays an important role in the applications which need privacy and authentication simultaneously. In order to get high efficiency in one-to-many communication environment and simplify the management of keys, a novel biometric signcryption scheme, identity-based and group-oriented, is proposed in this paper. In the scheme, the sender signcrypts the message and broadcasts it, but only the members of the receiver group can unsigncrypt the cipher text correctly. The implement of the scheme does not need bilinear pairing computation and modular exponentiation computation, hence it is very efficient. The analysis results show that the scheme is secure.

Cluster ensemble is becoming an important research pot and many researchers study in this field But there is no author to state the problem of distributed cluster ensemble In this paper the authors initiatively state the problem and introduce the model of distributed latent Dirichlet allocation (D-LDA) for distributed cluster ensemble which is the most important contribution of this paper First, the latent variables in D-LDA and some terminologies are defined for distributed cluster ensemble Second, Markov chain. Monte Carlo(MCMC) approximation inference for D-LDA is stated in detail Third, some datasets from UCI are chosen for experiments Compared with cluster-based similarity partitioning algorithm (CSPA), hyper-graph partitioning algorithm and meta-cluster algorithm (MCLA), the results show D-LDA does work better, further-more Me outputs of D-LDA, as a soft cluster model, Can not only cluster the, data ponds but also show the structure of data points

The management of private keys is a main problem in traditional threshold digital signature system. Combining a method of generating keys from biometric information and a modified RSA signature scheme, a novel biometric based (k, l) threshold signature scheme is proposed in this paper. The signer can sign a message only by his biometric information, and if the number of signers is equal to or more than k in a group of l members, the group signature can be generated. The security analysis results show that the scheme is secure and easy to implement.

The paper introduces a semi-supervised cluster ensemble of pairwised constrains based on the binary similarity matrix. Pairwised constrains are the typical way of semi-supervised learning. Cluster ensemble can increase robustness of clustering and it is helpful for knowledge reuse and distributed computing. The existing algorithms are mostly unsupervised algorithms of cluster ensemble which can't take advantages of known information ofdatasets. As a result the precision, robustness and stability of cluster ensemble are degraded. Semi-supervised cluster ensemble may conquer these disadvantages. The idea is that we use pairwised constrains as semi-supervised learning for semi-supervised cluster ensemble, in this paper there are three works presented. First, we state a semi-supervised cluster ensemble method. Second, the model of semi-supervised cluster ensemble is illustrated in detail. Third, some UCI datasets are chosen for the experiments, and the results show that semi-supervised cluster ensemble works well.