Search

As the most common idiopathic inflammatory myopathy in children, juvenile dermatomyositis (JDM) is characterized by skin rashes and muscle weakness. The childhood myositis assessment scale (CMAS) is commonly used to measure the degree of muscle involvement for diagnosis or rehabilitation monitoring. On the one hand, human diagnosis is not scalable and may be subject to personal bias. On the other hand, automatic action quality assessment (AQA) algorithms cannot guarantee 100% accuracy, making them not suitable for biomedical applications. As a solution, we propose a video-based augmented reality system for human-in-the-loop muscle strength assessment of children with JDM. We first propose an AQA algorithm for muscle strength assessment of JDM using contrastive regression trained by a JDM dataset. Our core insight is to visualize the AQA results as a virtual character facilitated by a 3D animation dataset, so that users can compare the real-world patient and the virtual character to understand and verify the AQA results. To allow effective comparisons, we propose a video-based augmented reality system. Given a feed, we adapt computer vision algorithms for scene understanding, evaluate the optimal way of augmenting the virtual character into the scene, and highlight important parts for effective human verification. The experimental results confirm the effectiveness of our AQA algorithm, and the results of the user study demonstrate that humans can more accurately and quickly assess the muscle strength of children using our system.

This paper proposes an improved three-stages cascading passivity-based control (PBC) for a grid -connected LCL converter in unbalanced weak grid condition. In general, the traditional double-loop control based on positive and negative sequence transformations is used in grid-connected converter control in unbalanced weak grid condition. However, it is time consuming for second harmonic filtering, and positive and negative sequence currents need to be controlled separately. The PBC has strong robustness to interference, and the line voltage based PBC can deal with the voltage unbalance effectively and easily without negative sequence transformation. But the traditional PBC needs six variables and three damping coefficients for the grid-connected LCL converter in unbalanced grid condition, and it has the disadvantage of difficult implementation. The improved PBC can realize the same control effect with three-stages cascading PBCs of two variables and one damping coefficient, and it has the advantages of easy implementation, good performance and high stability. First, the modeling and controller design are detailed described. Then the SIMULINK simulation results demonstrate the benefits of the improved control strategy. Finally, a grid-connected LCL converter prototype of 5kW is built and the experimental results verify the correctness and effectivity of the improved three-stages PBC strategy.

We introduce a unified system for SCHOLAT to import information into Encyclopedia, called Semi-automatic Scholar Encyclopedia Generating System, which consists of four modules: website display page, backstage administration module, version comparison and website security module. Scholar Encyclopedia is implemented as a subsystem of Semi-automatic Scholar Encyclopedia Generating System. Scholar Encyclopedia is a scholar information website that has been independently compiled by the scholars and filled in by personal information. Scholar Encyclopedia Import Model consists of three modules: user evaluation, information import and manual review. At present, Semi-automatic Scholar Encyclopedia Generating System has been used, and the beta version of Scholar Encyclopedia has been deployed in our school network center. The online application shows that Scholar Encyclopedia Import Model has good accuracy and practicability.

With the increasing tension of spectrum resources, millimeter wave communication has become a research hotspot. Carrier synchronization is one of the key technologies of millimeter wave communication. This paper analyzes in detail the carrier synchronization method of digital, analog, and analog-digital hybrid schemes. The digital carrier synchronization scheme has low transmission efficiency, and the analog carrier synchronization scheme has a complicated hardware structure. The digital-analog hybrid carrier synchronization scheme can achieve better synchronization performance under the condition of simpler hardware structure complexity.

Energy consumption, signal gain and spectral efficiency have become major concerns of 5G and millimeter wave, especially in the Internet of Things (IoT) scenario. Radiation pattern describes the dependence of the intensity and direction of a radio wave emitted by an antenna or other sources. The radiation pattern of the antenna array are easily affected by water molecules, dust, and the like in the air due to the densely packed antenna array in millimeter-wave system. The reflection and refraction of the antenna signal are caused by the bloakages, and the radiation pattern is changed. In this paper, a reduced model of the antenna diagnosis is built and the restructured iterative shrinkage-thresholding algorithm (ISTA-R) and restructured learning iterative shrinkage-thresholding algorithm (LISTA-R) are proposed to estimate the blocking coefficient and blocking position. The simulations show that the proposed algorithms can efficiently cut down the number of iterations and can improve the performance of real-time diagnosis.

Gallium nitride is becoming more popular in low-voltage applications. Gallium nitride (GaN) high electron mobility transistors (HEMTs) has positive temperature feature. It makes parallel application feasible for GaN HEMTs. Meanwhile, paralleled GaN HEMTs will increase the power handling capability and the efficiency of the converter. The parasitic parameters of both driving and power loop layout are critical which need to be equalized and minimized in GaN HEMT parallel operation. The parameters mismatch of the parallel branch will lead to significant effect on the dynamic characteristics and bring thermal problem of the GaN HEMTs. This paper focuses on the design of paralleled low-voltage enhancement GaN HEMT, and evaluates the effect of the parasitic inductances in both driver and power loops. The LT-spice and ANASYS Q3D Extractor simulation are carried out to analysis the effect of the unbalance parameters. The design guidelines for driver and PCB design are summarized as well. Finally, a 300W isolated DC-DC converter is built to verify the analysis and simulation on parallel operation.

With the development of broadcast video applications, the demand for broadcast video is increasing, which poses great challenges to the existing broadcast systems. The mmWave is capable of supporting large bandwidth. In addition, code domain NOMA is able to enhance the spectral efficiency. Therefore, combining the advantages of these two methods, the large-volume content services can be improved. This paper proposes to exploit the joint benefits of mmWave and code-domain NOMA to achieve further capacity gains in broadcast services. Theoretically analysis and numerical results validate the performance gain of the proposed scheme over conventional scheme with respect to spectral efficiency.

In this paper, extended state observer(ESO) is used to solve the static-state error caused by parasitic resistance in Buck converter with constant power load(CPL). Firstly, the control algorithm based on passivity-based control(PBC) is designed to solve the problem that the Buck converter with CPL can not operate stably, and the influence of parameter change on the control system is analyzed by using the zero-pole map. The influence caused by parasitic resistance is regarded as unknown disturbance, and it is estimated by ESO to reduce the control static-state error. Finally, the problem of non-convergence is solved by improving the fal function commonly used in ESO. Computer simulation shows the effectiveness and feasibility of the control strategy designed in this paper.

Deep neural networks are vulnerable to adversarial examples, which poses security concerns on these algorithms due to the potentially severe consequences. Adversarial attacks serve as an important surrogate to evaluate the robustness of deep learning models before they are deployed. However, most of existing adversarial attacks can only fool a black-box model with a low success rate. To address this issue, we propose a broad class of momentum-based iterative algorithms to boost adversarial attacks. By integrating the momentum term into the iterative process for attacks, our methods can stabilize update directions and escape from poor local maxima during the iterations, resulting in more transferable adversarial examples. To further improve the success rates for black-box attacks, we apply momentum iterative algorithms to an ensemble of models, and show that the adversarially trained models with a strong defense ability are also vulnerable to our black-box attacks. We hope that the proposed methods will serve as a benchmark for evaluating the robustness of various deep models and defense methods. With this method, we won the first places in NIPS 2017 Non-targeted Adversarial Attack and Targeted Adversarial Attack competitions., Comment: CVPR 2018 Spotlight

Visual question answering (VQA) requires joint comprehension of images and natural language questions, where many questions can't be directly or clearly answered from visual content but require reasoning from structured human knowledge with confirmation from visual content. This paper proposes visual knowledge memory network (VKMN) to address this issue, which seamlessly incorporates structured human knowledge and deep visual features into memory networks in an end-to-end learning framework. Comparing to existing methods for leveraging external knowledge for supporting VQA, this paper stresses more on two missing mechanisms. First is the mechanism for integrating visual contents with knowledge facts. VKMN handles this issue by embedding knowledge triples (subject, relation, target) and deep visual features jointly into the visual knowledge features. Second is the mechanism for handling multiple knowledge facts expanding from question and answer pairs. VKMN stores joint embedding using key-value pair structure in the memory networks so that it is easy to handle multiple facts. Experiments show that the proposed method achieves promising results on both VQA v1.0 and v2.0 benchmarks, while outperforms state-of-the-art methods on the knowledge-reasoning related questions., Comment: Supplementary to CVPR 2018 version

Integrated gate commutated thyristor (IGCT) shows excellent potential in high-power converters owing to its advantages of low on-state voltage, low cost, and high reliability. However, a di/dt limiting inductor and a clamp circuit must be added to IGCT converters to protect the freewheeling diodes (FWD) from severe reverse recovery, leading to additional size and losses. So far, the di/dt is still limited to lkA/μs in most of the IGCT converters, even the maximum turn-off di/dt of up-to-date fast recovery diodes has already reached up to 5.3kA/μs. In this paper, the maximum di/dt of both IGCT and diode is studied. The current commutation process of the clamp circuit is analyzed. The switching losses of IGCT, diode and clamp circuit are tested at different di/dt from 1.1kA/μs to 19.4kA/μs. Results show that the switching losses of IGCT converters will be reduced by 40% with lower di/dt limiting inductance. A back-to-back power cycling test of two half-bridge IGCT converters at 2200V/1500A and 150Hz verifies the long-term reliability of IGCT converters at maximum di/dt of 5.2kA/μs.

In order to adapt to the complicated and changeable topography environment features of the seabed, a composite wheeled submarine vehicle with both passive and passive hybrid obstacle negotiation model is proposed in this paper. Taking the composite wheeled seabed mining vehicle as the object of study, the features of seabed unstructured terrain environment are extracted, and the process of adaptive obstacle climbing is analyzed and simulated. The dynamic characteristic curve of seabed mining vehicle is obtained, which provides a reference for the development and test verification of seabed mining vehicle, and has important academic significance and engineering application value.

We propose BodyFusion, a novel real-time geometry fusion method that can track and reconstruct non-rigid surface motion of a human performance using a single consumer-grade depth camera. To reduce the ambiguities of the non-rigid deformation parameterization on the surface graph nodes, we take advantage of the internal articulated motion prior for human performance and contribute a skeleton-embedded surface fusion (SSF) method. The key feature of our method is that it jointly solves for both the skeleton and graph-node deformations based on information of the attachments between the skeleton and the graph nodes. The attachments are also updated frame by frame based on the fused surface geometry and the computed deformations. Overall, our method enables increasingly denoised, detailed, and complete surface reconstruction as well as the updating of the skeleton and attachments as the temporal depth frames are fused. Experimental results show that our method exhibits substantially improved nonrigid motion fusion performance and tracking robustness compared with previous state-of-the-art fusion methods. We also contribute a dataset for the quantitative evaluation of fusion-based dynamic scene reconstruction algorithms using a single depth camera.

This paper proposes a motion trajectory control method of underground intelligent scraper based on particle swarm optimization. The simulation model of whole system was established in simulink according to autonomous navigation bivariate PID control algorithm and scraper motion control model. The process of particle swarm optimization was designed based on MATLAB platform. The parameter optimization model of PID controller based on particle swarm optimization (PSO) algorithm is established. The parameters of PID controller are optimized, and the global position optimal solution is obtained. It obtained correlation curve of motion trajectory deviation through the experiment of motion trajectory of particle swarm optimization. The particle swarm optimization algorithm is proved to be very effective in controlling the trajectory of the intelligent scraper.

The deployment of deep convolutional neural networks (CNNs) in many real world applications is largely hindered by their high computational cost. In this paper, we propose a novel learning scheme for CNNs to simultaneously 1) reduce the model size; 2) decrease the run-time memory footprint; and 3) lower the number of computing operations, without compromising accuracy. This is achieved by enforcing channel-level sparsity in the network in a simple but effective way. Different from many existing approaches, the proposed method directly applies to modern CNN architectures, introduces minimum overhead to the training process, and requires no special software/hardware accelerators for the resulting models. We call our approach network slimming, which takes wide and large networks as input models, but during training insignificant channels are automatically identified and pruned afterwards, yielding thin and compact models with comparable accuracy. We empirically demonstrate the effectiveness of our approach with several state-of-the-art CNN models, including VGGNet, ResNet and DenseNet, on various image classification datasets. For VGGNet, a multi-pass version of network slimming gives a 20x reduction in model size and a 5x reduction in computing operations., Comment: Accepted by ICCV 2017

This paper presents a new type of electrically driven mining vehicle which adapts to the characteristics of seabed complex terrain environment. The optimal drive motor is selected and the corresponding control algorithm is designed. The obstacle surmounting process and the change law of the output torque of the electric driven mining vehicle under typical working conditions are analyzed. The simulation curves of the ground support force and the rolling resistance moment of the driving wheel in the process of obstacle surmounting are obtained, which can provide theoretical reference and technical support for the obstacle control strategy of the electric driven seabed mining vehicle. It has important academic significance and engineering application value.

Face recognition has been widely used in many application areas such as photo album management and information security. Rapid growth of handheld devices and social networks bring new challenges to face recognition algorithm design and system engineering. To be effective on a handheld device, the face recognition model must be simple and lightweight, and also needs to handle the large variations in background, image size, etc. This paper tries to address these two problems from a system perspective. We employ a simple linear model for face recognition, which is generated from a SVM classifier to tell whether a pair of face is from the same subject or not. And to make this classifier generalized well on all potential inter-face/intra-face variations, we train it with billions of face pairs. Many potential problems arise during the large-scale training procedure. For examples, 1) the training set is too large to fit into a single machine, 2) the computation requirement is so huge that it takes many days to complete a training which makes difficult for algorithm developers to experiment and tune the model effectively. To overcome these obstacles, this paper proposes a large-scale face recognition training system, called LFRTrainer. The system takes advantage of the massive data parallelism available in the problem and utilizes a distributed compute cluster to perform the model training. We have run this system on Intel Endeavor cluster for model development and tuning over a year. Using LFRTrainer, we train a face recognition model for embedded devices which achieves state-of-the-art accuracy and high scalability: 92.2% Face Verification Rate (FVR) at 0.1% False Accept Rate (FAR) on FRGC-204 benchmark and nearly linear scalability from tens to hundreds of nodes.

This paper presents the design method of omnidirectional Lamb-wave electromagnetic acoustic transducers (EMATs) for electromagnetic ultrasonic Lamb wave tomography technique. The 360 degrees radiated A0 mode Lamb wave is excited through the selection of the appropriate working point, which not only can realize large range detection of plate structure, but also be more convenient to extract feature of the echo. The travel time between the transmitting and receiving EMATs is measured through using crosshole scanning method. Then slowness image reconstruction is realized by using SIRT (Simultaneous Iterative Reconstruction Technique). The experimental result of imaging of a 1.5 mm deep artificial defect with a diameter of 26 mm in 2mm thick aluminum plate is obtained, according to which the defect can be found.

In a typical virtualized cloud system, network-attached storage (NAS) is the common technology for storing the disk image files of virtual machine (VM). Although the client/server model on NAS is more reliable and manageable, a concurrent accessing behavior from multiple VMs on the shared NAS server may exhaust the underlying network bandwidth, thus hurting the disk I/O performance of VMs. In this paper, we improve this problem by a set of optimizations, which are designed around bridging the semantic gap between the file system of VM and NAS server. The experimental result shows that our work can effectively reduce the network traffic between the NAS server and client. Meanwhile, the disk I/O performance of VMs can be also improved.

Social networks are very dynamic objects, since new edges and vertices are added to the graph over the time. Link prediction is an important task in social network analysis and is useful in many application domains. In the recent years, there is significant interest in methods that represent the social network in the form of a graph and leverage topological and semantic measures of similarity between two nodes to make predictions. In this article, we propose a hybrid approach utilizing time-varied weight information of links. We focus on the problem of link prediction particularly in the context of evolving co-authorship. Experiments have shown that the link prediction algorithm based on time-varied weight can reach better result.

IaaS cloud converts physical resources into resource pools. The pool oriented resource management model requires transparency of the underlying physical resources, as well as the on-demand consumption of IT services and the trust verification of the resources from cloud users. A viable way is by creating trusted clusters that consist of trusted physical machines, with strict membership management of these trusted clusters, confining IT services to be hosted within trusted clusters. This paper proposes a trusted cluster scheme which is based on a dynamic group signature scheme. The scheme fits into the pool oriented trust management for cloud computing.

Cloud computing enables rapid deployment of services on an on-demand basis in large scale. Public clouds are open to all users with identities from various domains, requiring the support of multiple identity providers and hybrid authentication protocols. The complexity of the authentication scenario brings a great burden to service providers. Service providers and users know only a small part of the complicated authentication network. This work explores a system view on the authentication network in cloud, allowing the authentication interaction and the trust relations to be explicitly described and analyzed. A cross-domain single sign-on model is proposed and the implementation details of the authentication architecture and protocol is presented.

Along with the inclusion of GPU cores within the same CPU die, the performance of Intel's processor-graphics has been significantly improved over earlier generation of integrated graphics. This paper presents a highly optimized SURF cascade based face detector which efficiently exploits both CPU and GPU computing power on the latest Sandy Bridge processor. The SURF cascade classifier procedure is partitioned into two phases in order to leverage both thread level and data level parallelism in the GPU. The integral image function running in the CPU core can work with the GPU in parallel. We measure the performance and power of the GPU implementation on the latest Sandy Bridge platform. The experimental results show that our proposed GPU implementation achieves a 2.98 speedup and a 1.42 speedup compared to the single thread and multi-thread CPU implementation. At the same time, the power usage can be reduced as much as 50% compared to the CPU implementation. In addition, our proposed method presents a general approach for task partitioning between CPU and GPU, thus being beneficial not only for face detection but also for other computer vision applications.

Collaboration networks are among some of the social networks and offer us the opportunity to study the structure underlying the networks. In this paper, we utilize the social network analysis (SNA) framework to understand what characterizes the social structure of the CSCWD conference's paper co-authorship network. It can potentially provide us with an understanding of the individuals and the network. We consider two scientists to be connected if they have authored a paper together, system like the co-authorship network of the conference is inherently dynamic, and so we represent it as a time-varying graph. Each graph is a static cumulative network. We then give results for the average distance and the diameter they show that this network forms a “small world”. We then also give the clustering coefficient demonstrating the presence of clustering of the collaboration between the scientists, as well as the increasing proportion of the giant component, it can serve as an additional support that the community would work well if it is densely connected. Through those figures, we also want to find some potential actions that can be taken to further develop the conference and other similar conferences.

We present a novel boosting cascade based face detection framework using SURF features. The framework is derived from the well-known Viola-Jones (VJ) framework but distinguished by two key contributions. First, the proposed framework deals with only several hundreds of multidimensional local SURF patches instead of hundreds of thousands of single dimensional haar features in the VJ framework. Second, it takes AUC as a single criterion for the convergence test of each cascade stage rather than the two conflicting criteria (false-positive-rate and detection-rate) in the VJ framework. These modifications yield much faster training convergence and much fewer stages in the final cascade. We made experiments on training face detector from large scale database. Results shows that the proposed method is able to train face detectors within one hour through scanning billions of negative samples on current personal computers. Furthermore, the built detector is comparable to the state-of-the-art algorithm not only on the accuracy but also on the processing speed.

As a new generation of high-capacity Voltage Source Converter (VSC), Modular Multilevel Converter (MMC) is a potential topology of VSC based High Voltage Direct Current Transmission System (VSC-HVDC). This paper focuses on the voltage balancing control of floating dc capacitors based on CPS-PWM. The imbalance mechanics of capacitor voltage between and within arms are theoretically revealed. A new balancing control strategy called Capacitor Voltage Balancing Control Loop (CVBC Loop) is proposed, which can directly and accurately regulate the active power injected into the MMC module to guarantee dc voltage balance of capacitors within the same arm. A 10kVA laboratory prototype has been built. The experiment results verify the validity of the proposed control strategy under both steady and dynamic state.

Facial landmark detection is an essential module in many face related applications and it often appears as the most time consuming part in face processing pipeline. This paper proposes a fast and effective method for facial landmark detection using Haar cascade classifiers and a simple 3D head model, which not only detects the position of landmark points but also gives an estimation of head pose such as yaw and pitch angles. To reduce the amount of computation, only 7 landmark points are detected (4 eye corners, 2 mouth corners, 1 nose tip) that generally meets the requirement of face alignment and face recognition. Experiment on multiple datasets shows our algorithm can provide sufficient accuracy of facial landmark localization while being able to run in super real-time at Intel Atom 1.3 GHz embedded processors.

This essay aims at solving the problems such as low efficiency, high-error rates and inadequate real-time data acquisitions when using a portable soil electrical conductivity meter. Based on the theory of multi-sensor data fusion, a new method of measurement of soil electrical conductivity is given, with which soil electrical conductivity can be measured with the single sensor fusion algorithm of batch estimation combining with the method of adaptive weighted fusion. At the same time, a hardware system that achieves the algorithm is designed, and the soil electrical conductivity can be collected in real time by the hardware system. Thus, the problems can be solved by using the new measurement method. Innovation point of the essay is the method that the soil electrical conductivity can measured by the single sensor fusion algorithm of batch estimation combining with the method of adaptive weighted fusion.

A vision-aided inertial navigation algorithm is proposed within an EKF framework for spacecraft precision entry, descent and landing on Mars. By combing vision information with inertial measurement, it can not only avoid error accumulation of the IMU dead reckoning, but also obtain accurate and robust sate estimation in a real-time and high rate fashion. A special procedure based on state augmentation is developed to account for the delays of image processing. The problem associated with the sparse measurement information during the descent terminal phase is addressed as well by processing the previously unknown features. Simulation results demonstrate that the proposed integrated navigation system can provide high-precision motion information required by the upcoming planetary exploration missions.

A NASICON-based electrochemical device attached with an oxide electrode was developed for the detection of Cl 2 in air at high temperature. Among the oxides tested, Cr 2 O 3 was found to be the best suited for the sensing electrode. The ΔEMF value of the device attached with Cr 2 O 3 was almost linear with the logarithm of 5–50 ppm Cl 2 concentration. It was also observed that the sensor showed a speedy response and recovery kinetics to Cl 2 . In addition, the device exhibited excellent selectivity to Cl 2 against the other interference gases, such as CO, NO, NO 2 , H 2 , C 7 H 8 , C 2 H 4 and CH 4 . The buried reference electrode can effectively improve the sensitivity of the sensor.

This paper focuses on the gas sensing properties of the mixed-potential-type yttria stabilized zirconia (YSZ)-based NO 2 sensor with NiO sensing electrode. The sensing characteristics of the sensors have been improved by controlling the microstructure of NiO sensing electrode synthesized with the microwave. The microstructure of NiO strongly depended on the microwave treatment temperature and the sintering temperature. The largest sensitivity of the sensor was obtained by using the NiO synthesized at 60°C and sintered at 1200°C, the slope of the dependence of EMF on the NO 2 concentration in the range of 20–500 ppm was 54 mV/decade at 850°C. It was also seen that the sensor showed a speedy response and recovery kinetics to NO 2 . The sensing mechanism was confirmed to involve in the microstructure of oxide sensing electrode.

Solid state electrochemical sensors using yttria-stabilized zirconia (YSZ) plate and Cr 2 O 3 -WO 3 nanocomposites were fabricated. The effects of the sintering temperature on the electrode microstructure were investigated. It was found that the sensor sintered at 1000 °C exhibited larger sensitivity to NO 2 in the examined range of 20–300ppm than other sensors sintered at 800 °C, 900 °C, and 1100 °C; and it was speculated that the microstructure of the SE was the primary factor. The NO 2 sensing characteristics of the sensors were examined at high temperature.

As the Internet becomes a common tool in people's daily life, more and more information is acquired through webpage. However, the website security has became an important issue in recent years. The tamper-proof of webpage is of great importance. Some watermarking schemes have been reported to solve this problem. For a webpage, we create the watermark according to the webpage tags by using SVD, and then embed them into the webpage If the extracted watermark from the received webpages is not the same as original watermark, then it shows that the webpage has been modified, and we can also locate the modified position. Experiments are performed on the proposed scheme and the results show that the proposed scheme can be a feasible and efficient tool for the tamper-proof of webpage. So our scheme plays an important role in the webpage authentication.

This paper presents a general texture mapping framework for image-based 3D modeling. It aims to generating seamless texture map for 3D model created by real-world photos under uncontrolled environment. Our proposed method addresses two challenging problems: 1) texture discontinuity due to system error in 3D modeling from self-calibration; 2) color/lighting difference among images due to real-world uncontrolled environments. The general framework contains two stages to resolve these problems. The first stage globally optimizes the registration of texture patches and triangle faces with Markov Random Field (MRF) to optimize texture mosaic. The second stage does local radiometric correction to adjust color difference between texture patches and then blend texture boundaries to improve color continuity. The proposed method is evaluated on several 3D models by image-based 3D modeling, and demonstrates promising results.