The facial expression is diverse and various among persons due to the impact of the psychology factor. Whilst the facial action is comparatively steady because of the fixedness of the anatomic structure. Therefore, to improve performance of the action unit recognition will facilitate the facial expression recognition and provide profound basis for the mental state analysis, etc. However, it still a challenge job and recognition accuracy rate is limited, because the muscle movements around the face are tiny and the facial actions are not obvious accordingly. Taking account of the moving of muscles impact each other when person express their emotion, we propose to make full use of co-occurrence relationship among action units (AUs) in this paper. Considering the dynamic characteristic of AUs as well, we adopt the 3D Convolutional Neural Network(3DCNN) as base framework and proposed to recognize multiple action units around brows, nose and mouth specially contributing in the emotion expression with putting their co-occurrence relationships as constrain. The experiments have been conducted on a typical public dataset CASME and its variant CASME2 dataset. The experiment results show that our proposed AU co-occurrence constraint 3DCNN based AU recognition approach outperforms current approaches and demonstrate the effectiveness of taking use of AUs relationship in AU recognition.

With the advent of IoT technology and Big Data computing, the importance of WSNs (Wireless Sensor Networks) has been on the rise. For energy-efficient and collection-efficient delivery of any sensed data, lots of novel wireless medium access control (MAC) protocols have been proposed and these MAC schemes are the basis of many IoT systems that leads the upcoming fourth industrial revolution. WSNs play a very important role in collecting Big Data from various IoT sensors. Also, due to the limited amount of battery driving the sensors, energy-saving MAC technologies have been recently studied. In addition, as new IoT technologies for Big Data computing emerge to meet different needs, both sensors and sinks need to be mobile. To guarantee stability of WSNs with dynamic topologies as well as frequent physical changes, the existing MAC schemes must be tuned for better adapting to the new WSN environment which includes energy-efficiency and collection-efficiency of sensors, coverage of WSNs and data collecting methods of sinks. To address these issues, in this paper, a self-organization scheme for mobile sensor networks with mobile multiple sinks has been proposed and verified to adapt both mobile sensors and multiple sinks to 3-dimensional group management MAC protocol. Performance evaluations show that the proposed scheme outperforms the previous schemes in terms of the various usage cases. Therefore, the proposed self-organization scheme might be adaptable for various computing and networking environments with big data.

The accurate segmentation of infant brain MR image into white matter (WM), gray matter (GM), and cerebrospinal fluid (CSF) is very important for early studying of brain growing patterns and morphological changes in neurodevelopmental disorders. Because of inherent myelination and maturation process, the WM and GM of babies (between 6 and 9 months of age) exhibit similar intensity levels in both T1-weighted (T1w) and T2-weighted (T2w) MR images in the isointense phase, which makes brain tissue segmentation very difficult. We propose a deep network architecture based on U-Net, called Triple Residual Multiscale Fully Convolutional Network (TRMFCN), whose structure exists three gates of input and inserts two blocks: residual multiscale block and concatenate block. We solved some difficulties and completed the segmentation task with the model. Our model outperforms the U-Net and some cutting-edge deep networks based on U-Net in evaluation of WM, GM and CSF. The data set we used for training and testing comes from iSeg-2017 challenge (http://iseg2017.web.unc.edu).

The k-means algorithm is widely used in academia and industry due to easy and simple implementation, enabling fast learning for complex datasets. However, k-means struggles to classify datasets without prior knowledge of specific domains. We proposed the repulsive k-means (RK-means) algorithm in a previous study to improve the k-means algorithm, using the repulsive force concept, which allows deleting unnecessary cluster centroids. Accordingly, the RK-means enables to classifying of a dataset without domain knowledge. However, three main problems remain. The RK-means algorithm includes a cluster repulsive force offset, for clusters confined in other clusters, which can cause cluster locking; we were unable to prove RK-means provided optimal convergence in the previous study; and RK-means shown better performance only normalize term and weight. Therefore, this paper proposes the advanced RK-means (ARK-means) algorithm to resolve the RK-means problems. We establish an initialization strategy for deploying cluster centroids and define a metric for the ARK-means algorithm. Finally, we redefine the mass and normalize terms to close to the general dataset. We show ARK-means feasibility experimentally using blob and iris datasets. Experiment results verify the proposed ARK-means algorithm provides better performance than k-means, k'-means, and RK-means.

E-health services have provided interoperability between personal health devices in personal area network, based the ISO/IEEE 11073 standard. In the healthcare system, the manager handles most agents concurrently through wireless communication. However, due to the distance limitation and the increased number of agents, it may be difficult to provide continuous connectivity. Recently, body area devices have been equipped with various applicable agents, which can even handle agents on behalf of the manager. A BAD may act as an intermediary device to increase system efficiency and performance. In this study, a device called "proxy", which can be installed as software on BAD devices, is proposed. The data measured by an agent can be sent to the proxy first, and subsequently be sent to the manager again. Agents and the manager are not aware of the proxy existence and work normally without the proxy. Furthermore, a new smart proxy and modified manager are proposed. The smart proxy acts as one agent handling measurement data from several agents, which can transmit a significant amount of data at once. The proxy and smart proxy maintain compatibility with existing devices that conform to the 20601 standard. The proposed schemes are verified and the complexities of devices are analyzed. The analysis shows no significant difference among the proxy, smart proxy, and manager. Simulations exhibit that the proposed schemes can improve the system performance.

Cloud computing services changed the way the data are managed across the healthcare system that can improve patient care. Currently, most healthcare organizations are using cloud-based applications and related services to deliver better healthcare facilities. But architecting a cloud-based healthcare system needs deep knowledge about the working nature of these services and the requirements of the healthcare environment. The success is based on the usage of appropriate cloud services in the architecture to manage the data flow across the healthcare system.Cloud service providers offer a wide variety of services to ingest, store and process healthcare data securely. The top three public cloud providers- Amazon, Google, and Microsoft offers advanced cloud services for the solution that the healthcare industry is looking for. This article proposes a framework that can effectively utilize cloud services to handle the data flow among the various stages of the healthcare infrastructure. The useful cloud services for ingesting, storing and analyzing the healthcare data for the proposed framework, from the top three cloud providers are listed in this work. Finally, a cloud-based healthcare architecture using Amazon Cloud Services is constructed for reference.

People are increasingly demanding to experience realistic behavior of virtual characters in computer games. In this article, we build a computer model for simulating a bicycle rider's behavior in a virtual riding system. A method to calculate the speed of a bicycle rider is proposed to improve the reality in a virtual riding system. In this method, the property of physical energy is introduced, and the bicycle-riding speed is calculated in real time according to the relationship between the rider's physical energy and bicycle-riding speed. Then based on the analysis of the behavior of a cyclist in a real competition, various behaviors of the virtual rider are designed and a behavior-tree for the virtual bicycle rider is constructed accordingly. On the basis of these, a virtual riding system is developed. The experiments results show that our system can simulate the behavior of a virtual bicycle rider, and thus encourage exercise on a stationary bicycle.

A kind of Subjective Imaging Effect Assessment (SIEA) method and its applications on intelligent imaging terminal design in engineering site are presented. First, some visual assessment indices are used to characterize the imaging effect: the image brightness, the image brightness uniformity, the color image contrast, the image edge blur, the image color difference, the image saturation, the image noise, and the integrated imaging effect index. A linear weighted function is employed to carry out the SIEA computation and the Analytic Hierarchy Process (AHP) technique is used to estimate its weights. Second, a SIEA software is developed. It can play images after the settings of assessment index or assessment reaction time, etc. Third, two cases are used to illustrate the application effects of proposed method: the image enhancement system design for surveillance camera and the imaging environment perception system design for intelligent lighting terminal. A Prior Sequential Stimulus (PSS) experiment is proposed to improve the evaluation stability of SIEA method. Many experiment results have shown the proposed method can realize a stable system design or parameters setting for the intelligent imaging terminal in engineering site.

With rapid growth of content demands, device-to-device (D2D) content sharing is exploited to effectively improve the service quality of users. Considering the limited storage space and various content demands of users, caching schemes are significant. However, most of them ignore the influence of the asynchronous content reuse and the selfishness of users. In this work, the user preferences are defined by exploiting the user-oriented content popularity and the current caching situation, and further, we propose the social-aware rate, which comprehensively reflects the achievable contents download rate affected by the social ties, the caching indicators, and the user preferences. Guided by this, we model the collaborative caching problem by making a trade-off between the redundancy of caching contents and the cache hit ratio, with the goal of maximizing the sum of social-aware rate over the constraint of limited storage space. Due to its intractability, it is computationally reduced to the maximization of a monotone submodular function, subject to a matroid constraint. Subsequently, two social-aware collaborative caching algorithms are designed by leveraging the standard and continuous greedy algorithms respectively, which are proved to achieve different approximation ratios in unequal polynomial-time. We present the simulation results to illustrate the performance of our schemes.

The automatic identification and classification of image-based weld defects is a difficult task due to the complex texture of the X-ray images of the weld defect. Several depth learning methods for automatically identifying welds were proposed and tested. In this work, four different depth convolutional neural networks were evaluated and compared on the 1631 image set. The concavity, undercut, bar defects, circular defects, unfused defects and incomplete penetration in the weld image 6 different types of defects are classified. Another contribution of this paper is to train a CNN model "RayNet" for the dataset from scratch. In the experiment part, the parameters of convolution operation are compared and analyzed, in which the experimental part performs a comparative analysis of various parameters in the convolution operation, compares the size of the input image, gives the classification results for each defect, and finally shows the partial feature map during feature extraction with the classification accuracy reaching 96.5%, which is 6.6% higher than the classification accuracy of other existing fine-tuned models, and even improves the classification accuracy compared with the traditional image processing methods, and also proves that the model trained from scratch also has a good performance on small-scale data sets. Our proposed method can assist the evaluators in classifying pipeline welding defects.

The chest X-rays are a common way to diagnose lung cancer or pneumonia. In particular, the finding of a lung nodule is the most important problem in the early detection of lung cancer. Recently, a lot of automatic diagnosis algorithms have been studied to find the lung nodules missed by doctors. The algorithms are typically based on segmentation network like U-Net. However, the occurrence of false positives that similar to lung nodules present outside the lungs can severely degrade performance. In this study, we propose a multi-task learning method that simultaneously learns the lung region and nodule-labeled data based on the prior knowledge that lung nodules exist only in the lung. The proposed method significantly reduces false positives outside the lung and improves the recognition rate of lung nodules to 83.8 F1 score compared to 66.6 F1 score of single task learning with U-net model. The experimental results on the JSRT public dataset demonstrate the effectiveness of the proposed method compared with other baseline methods.

We propose a robust visual object tracking algorithm fusing a convolutional neural network tracker trained offline from a large number of video repositories and a color histogram based tracker to track objects for mixing immersive audio. Our algorithm addresses the problem of occlusion and large movements of the CNN based GOTURN generic object tracker. The key idea is the offline training of a binary classifier with the color histogram similarity values estimated via both trackers used in this method to opt appropriate tracker for target tracking and update both trackers with the predicted bounding box position of the target to continue tracking. Furthermore, a histogram similarity constraint is applied before updating the trackers to maximize the tracking accuracy. Finally, we compute the depth(z) of the target object by one of the prominent unsupervised monocular depth estimation algorithms to ensure the necessary 3D position of the tracked object to mix the immersive audio into that object. Our proposed algorithm demonstrates about 2% improved accuracy over the outperforming GOTURN algorithm in the existing VOT2014 tracking benchmark. Additionally, our tracker also works well to track multiple objects utilizing the concept of single object tracker but no demonstrations on any MOT benchmark.

Salp Swarm Algorithm (SSA) is a new nature-inspired swarm optimization algorithm that mimics the swarming behavior of salps navigating and foraging in the oceans. SSA has been proved to enable to avoid local optima and enhance convergence speed benefiting from the adaptive nonlinear mechanism and salp chains. In this paper, visual tracking is considered to be a process of locating the optimal position through the interaction between leaders and followers in successive images. A novel SSA-based tracking framework is proposed and the analysis and adjustment of parameters are discussed experimentally. Besides, the qualitative analysis and quantitative analysis are performed to demonstrate the tracking effect of our proposed approach by comparing with ten classical tracking algorithms. Extensive comparative experimental results show that our algorithm has good performance in visual tracking, especially for abrupt motion tracking.

With the development of medical imaging technology, image registration has been widely used in the field of disease diagnosis. The registration between different modal images of brain magnetic resonance (MR) is particularly important for the diagnosis of brain diseases. However, previous registration methods don't take advantage of the prior knowledge of bilateral brain symmetry. Moreover, the difference in gray scale information of different modal images increases the difficulty of registration. In this paper, a multimodal medical image registration method based on image segmentation and symmetric self-similarity is proposed. This method uses modal independent self-similar information and modal consistency information to register images. More particularly, we propose two novel symmetric self-similarity constraint operators to constrain the segmented medical images and convert each modal medical image into a unified modal for multimodal image registration. The experimental results show that the proposed method can effectively reduce the error rate of brain MR multimodal medical image registration with rotation and translation transformations (average 0.43mm and 0.60mm) respectively, whose accuracy is better compared to state-of-the-art image registration methods.

The human expression contains a lot of information that can be used to detect complex conditions such as pain and fatigue. After deep learning became the mainstream method, the traditional feature extraction method no longer has advantages. However, in order to achieve higher accuracy, researchers continue to stack the number of layers of the neural network, which makes the real-time performance of the model weak. Therefore, this paper proposed an expression recognition framework based on densely concatenated convolutional neural networks to balance accuracy and latency and apply it to humanoid robots. The techniques of feature reuse and parameter compression in the framework improved the learning ability of the model and greatly reduced the parameters. Experiments showed that the proposed model can reduce tens of times the parameters at the expense of little accuracy.

The tradeoff between energy conservation and traffic balancing is a dilemma problem in Wireless Sensor Networks (WSNs). By analyzing the intrinsic relationship between cluster properties and long distance transmission energy consumption, we characterize three node sets of the cluster as a theoretical foundation to enhance high performance of WSNs, and propose optimal solutions by introducing rendezvous and Mobile Elements (MEs) to optimize energy consumption for prolonging the lifetime of WSNs. First, we exploit an approximate method based on the transmission distance from the different node to an ME to select suboptimal Rendezvous Point (RP) on the trajectory for ME to collect data. Then, we define data transmission routing sequence and model rendezvous planning for the cluster. In order to achieve optimization of energy consumption, we specifically apply the economic theory called Diminishing Marginal Utility Rule (DMUR) and create the utility function with regard to energy to develop an adaptive energy consumption optimization framework to achieve energy efficiency for data collection. At last, Rendezvous Transmission Algorithm (RTA) is proposed to better tradeoff between energy conservation and traffic balancing. Furthermore, via collaborations among multiple MEs, we design Two-Orbit Back-Propagation Algorithm (TOBPA) which concurrently handles load imbalance phenomenon to improve the efficiency of data collection. The simulation results show that our solutions can improve energy efficiency of the whole network and reduce the energy consumption of sensor nodes, which in turn prolong the lifetime of WSNs.

Steganalysis & steganography have witnessed immense progress over the past few years by the advancement of deep convolutional neural networks (DCNN). In this paper, we analyzed current research states from the latest image steganography and steganalysis frameworks based on deep learning. Our objective is to provide for future researchers the work being done on deep learning-based image steganography & steganalysis and highlights the strengths and weakness of existing up-to-date techniques. The result of this study opens new approaches for upcoming research and may serve as source of hypothesis for further significant research on deep learning-based image steganography and steganalysis. Finally, technical challenges of current methods and several promising directions on deep learning steganography and steganalysis are suggested to illustrate how these challenges can be transferred into prolific future research avenues.

Along with the increase of the importance of information used in practice, adversaries tried to take valuable information in diverse ways. The simple and fundamental solution is to encrypt the whole data. Since the cost of encryption is increasing along with the size of data, the cost for securing the data is a burden to a system where the size of the data is not small. For the reason, in some applications where huge data are used for service, service providers do not use any encryption scheme for higher security, which could be a source of trouble. In this work, we introduce a new type of data securing technique named Trapdoor Digital Shredder(TDS) which disintegrates a data to multiple pieces to make it hard to re-construct the original data except the owner of the file who holds some secret keys. The main contribution of the technique is to increase the difficulty in obtaining private information even if an adversary obtains some shredded pieces. To prove the security of our scheme, we first introduce a new security model so called IND-CDA to examine the indistinguishability of shredded pieces. Then, we show that our scheme is secure under IND-CDA model, which implies that an adversary cannot distinguish a subset of shreds of a file from a set of random shreds.

Activating appropriate roles for a session in the role-based access control (RBAC) model has become challenging because of the so-called role explosion. In this paper, factors and issues related to user-driven role management are analysed, and a session role activation (SRA) problem based on reasonable assumptions is proposed to describe the problem of such role management. To solve the SRA problem, we propose an extended RBAC model with context-based role filtering. When a session is created, context conditions are used to filter roles that do not need to be activated for the session. This significantly reduces the candidate roles that need to be reviewed by the user, and aids the user in rapidly activating the appropriate roles. Simulations are carried out, and the results show that the extended RBAC model is effective in filtering the roles that are unnecessary for a session by using predefined context conditions. The extended RBAC model is also implemented in the Apache Shiro framework, and the modifications to Shiro are described in detail.

In the modern financial sector, interest in providing financial services that employ blockchain technology has increased. Blockchain technology is efficient and can operate without a trusted party to store all transaction information; additionally, it provides transparency and prevents the tampering of transaction information. However, new security threats can occur because blockchain technology shares all the transaction information. Furthermore, studies have reported that the private keys of users who use the same signature value two or more times can be recovered. Because private keys of blockchain identify users, private key leaks can result in attackers stealing the ownership rights to users' property. Therefore, as more financial services use blockchain technology, actions to counteract the threat of private key recovery must be continually investigated. Private key recovery studies are presented here. Based on these studies, duplicated signatures generated by blockchain users are defined. Additionally, scenarios that generate and use duplicated signatures are applied in an actual bitcoin environment to demonstrate that actual bitcoin users' private keys can be recovered.

Using decode-and-forward relaying in the cognitive radio networks, the spectrum efficiency can improve furthermore. The optimization algorithm of the spectrum sensing estimation time is presented for the cognitive relay networks in this paper. The longer sensing time will bring two aspects of the consequences. On the one hand, the channel parameters are estimated more accurate so as to reduce the interferences to the authorized users and to improve the throughput of the cognitive users. On the other hand, it shortens the transmission time so as to decease the system throughput. In this time, it exists an optimal sensing time to maximize the throughput. The channel state information of the sub-bands is considered as the exponentially distributed, so a stochastic programming method is proposed to optimize the sensing time for the cognitive relay networks. The computer simulation results using the Matlab software show that the algorithm is effective, which has a certain engineering application value.

Cognitive radio (CR) is regarded as an effective approach to avoid the inefficient use of spectrum. However, CRNs have more special security problems compared with the traditional wireless communication systems due to its open and dynamic characteristics. Primary user emulation attack (PUEA) is a common method which can hinder secondary users (SUs) from accessing the spectrum by transmitting signals who has the similar characteristics of the primary users' (PUs) signals, and then the SUs' quality of service (QoS) cannot be guaranteed. To handle this issue, we first design a multiple-phase energy detection scheme based on the cooperation of multiple SUs to detect the PUEA more precisely. Second, a joint SUs scheduling and power allocation scheme is proposed to maximize the weighted effective capacity of multiple SUs with a constraint of the average interference to the PU. The simulation results show that the proposed method can effectively improve the effective capacity of the secondary users compared with the traditional overlay scheme which cannot be aware of the existence of PUEA. Also the good delay QoS guarantee for the secondary users is provided.

Advances in localization-based technologies and the increase in ubiquitous computing have led to a growing interest in location-based applications and services. High accuracy of the position of a wireless device is still a crucial requirement to be satisfied. Firstly, the rapid development of wireless communication technologies has affected the location accuracy of radio monitoring systems employed locally and globally. Secondly, the location is determined using standard complex computing methods and needs a relatively long execution time. In this paper, two geolocalization techniques, based on trigonometric and CORDIC computing processes, are proposed and implemented for Bluetooth-based indoor monitoring applications. Theoretical analysis and simulation results are investigated in terms of accuracy, scalability, and responsiveness. They show that the proposed techniques can locate a target wireless device accurately and are well suited for timing estimation.

An improved clustering scheme based on user group is proposed. Every two users are grouped among N-users in the allowed system according to their link gain from large to small. Each user group is numbered sequentially. Two user clusters are obtained according to the principle of maximizing link gain difference for the users in the first and last user groups. The remaining user groups are added to the two existing user clusters according to the parity of the group number. The clustering should be clustered again among the users in either user cluster if the throughput summation of a user cluster in NOMA is less than that of these users in orthogonal multiple access. The simulation results show that the proposed clustering scheme can increase the system throughput by about 8% compared with the hybrid clustering scheme when the number of users requiring service is 12.

In search and rescue mission, micro aerial vehicles (MAVs) are typically used to capture image and video from an aerial perspective and transfer the data to the ground station. Because of the power limitation, a cluster of MAVs are required for a large search area, hence an ad-hoc wireless network must be maintained to transfer data more conveniently and fast. However, the unstable link and the intermittent connectivity between the MAVs caused by MAVs' movement may challenge the packet forwarding. This paper proposes a delay tolerant packet forwarding algorithm based on location estimation for MAV networks, called DTN_est algorithm. In the algorithm, ferrying MAVs are used to transmit data between MAVs and the ground station, and the locations of both searching MAVs and ferrying MAVs are estimated to compute the distances between the MAVs and destination. The MAV that is closest to the destination is selected greedy to forward packet. If a MAV cannot find the next hop MAV using the greedy strategy, the packets will be stored and re-forwarded once again in the next time slot. The experiment results show that the proposed DTN_est algorithm outperforms the typical DTN_geo algorithm in terms of packet delivery ratio and average routing hops.