Search

Low-power consumption of orientation estimation using low-cost inertial sensors are crucial for all the applications which are resource constrained critically. This paper presents a novel Lightweight quaternion-based Extended Kalman Filter (LEKF) for orientation estimation for magnetic, angular rate and gravity (MARG) sensors. In this filter, with employing the quaternion kinematic equation as the process model, we derived a simplified measurement model to create the lightweight system model for Kalman filtering, where the measurement model works efficiently, and the involved computation of measurement model is reduced. It’s later proved that the proposed filter saves time consumption. Further, observations of the accelerometer and the magnetometer are dealt with the proposed nonlinear measurement model, the good performance should be guaranteed in theory. For the experiments, a commercial sensor for data collection, and an optical system to provide reference measurements of orientation, namely Vicon, are utilized to investigate the performance of the proposed filter. Evaluation for different application scenarios are considered, which primarily includes human motion capture and the drone application. Results indicate that the proposed filter performs well for both applications, and is proved to have a good anti-vibration ability. What’s more, the comparison experiment shows that the proposed filter provides good performance, as same as the basic UKF, but it consumes the least computing time among the compared methods.

This paper addresses an open vehicle routing problem with predetermined time windows under uncertain travel times (OVRP-UT). A novel light-robust-optimization model is proposed by integrating the goal programming formulations with set-based descriptions of the problem data, which can enable as many customers as possible to meet their demands within a group of predetermined time windows. An effective memetic algorithm (MA) is presented for solving the OVRP-UT model. We design a heuristic-based initialization mechanism to generate an initial population with a high level of quality and diversity. We design a timely-vertices based crossover operator and mutation operator to give birth to the offspring with high quality and good structure built in the search process. We provide a hybrid selection mechanism and a population updating strategy to remain the diversity of the population. We develop a self-adapted crossover and mutation rate to help the MA suit the different phases during the search process. A comprehensive simulation experiment based on the 320 benchmark instances demonstrates the effectiveness of the proposed algorithm.

In skeleton-based human action recognition domain, the methods based on graph convolution networks have great success recently. However, most graphical neural networks consider the skeleton as a spatiotemporally uncorrelated graph and rely on a predetermined adjacency matrix, ignoring the spatiotemporal relevance of human actions and taking up significant computational costs. Meanwhile, the methods use graph convolution to focus too much on the neighboring nodes of the joints and ignore the totality of the action. In this work, we propose a lightweight but efficient neural network called NLB-ACSE based on the Graph Convolutional Network (GCN). Our model consists of two large branches: non-local block branch that focuses on the long distance features and adaptive cross-spacetime edge branch that focuses on the short distance features. Both branches extract information across time and space, and focus on long and short information. Some simple but effective strategies also are applied to our model, such as semantics, maxpooling, and fusion inputs, which have small parameter burden but obtain a higher accuracy on ablation study. The proposed method with an order of magnitude smaller size than most previous papers is evaluated on three large datasets, NTU60, NTU120, and Northwesten-UCLA. The experimental results show that our method achieves the state-of-the-art performance.

In this paper, the influence of data imbalance on neural networks is discussed, and an improved learning algorithm to solve this problem is proposed. The experimental results show that in the case of imbalanced data, the training error of neural network converges slowly and the generalization ability is poor. Our theoretical analysis shows that in the process of training, the gradient descent direction of the weights is dominated by the major-classes, which accounts for the slow convergence of the training error. Based on these results, we propose the Equilibration Mini-batch Stochastic Gradient Descent (EMSGD) method to ensure the equilibrium of the data in the mini-batch. The advantage of this technique is that it makes full use of the existing random sampling step of MSGD without increasing the computational complexity. In addition, by over-sampling of minor-classes in the mini-batch, duplicated instances would be greatly reduced, thus preventing the model from overfitting. The experimental results show that under the condition of the imbalanced training data, EMSGD can make the neural network training error converge rapidly.

This article describes an improved behavior model for IGBT modules. The new steady-state characteristic model of the IGBT with a correction function was built to improve the saturation current for different gate-emitter voltages. Two important nonlinear capacitances were described by two continuous functions to solve the convergence problem in the circuit simulation. The parasitic parameters generated by the package were extracted by the impedance measurement. All the parameters of the model are derived from datasheet and measurement. The improved model was implemented by MAST language in SABER circuit simulation. A double-pulse test was applied to verify the improved model. Compared with the conventional model for transient waveforms, switching losses, and electromagnetic interference (EMI), the improved model shows a better agreement with experiments. The satisfactory results indicate that the improved model can offer a simple, fast, and generic modeling approach.

The practical application of 3D inversion of gravity data requires a lot of computation time and storage space. To solve this problem, we present an integrated optimization algorithm with the following components: (1) targeting high accuracy in the space domain and fast computation in the wavenumber domain, we design a fast 3D forward algorithm with high precision; and (2) taking advantage of the symmetry of the inversion matrix, the main calculation in gravity conjugate gradient inversion is decomposed into two forward calculations, thus optimizing the computational efficiency of 3D gravity inversion. We verify the calculation accuracy and efficiency of the optimization algorithm by testing various grid-number models through numerical simulation experiments.

With regards to use Virtual Reality headsets, people have many opportunities of key entry. However, current methods for key entry are suffering from low input speed and high error rate. In this paper, we suggest a new key entry system, Air keyboard with haptic finger glove. By applying haptic illusions such as phantom sensation and apparent movement, the finger glove could simulate the tactile feedback such as typing and unevenness sensation with low cost. Several controlled experiments were conducted to evaluate the efficiency and the accuracy of the air keyboard together with haptic finger glove. The result shows that the haptic finger glove could help user to better interact with the virtual keyboard. Moreover, a short-term training could increase the input efficiency and reduce the error rate, which is comparable or better than traditional pointing method with handheld controllers.

In recent years, a variety of oral care products have been developed. In this research, we propose a toothbrushing system that can detect force from the force-sensitive sensor. The results of the experiments show that the membrane force sensors attached to the brush side have a higher detection rate of force than the acceleration sensor.

Due to the COVID-19 pandemic, remote collaboration becomes widely needed. To this end, this research proposes a VR whiteboard system that facilitates multiple participants collaborate remotely using natural handwriting. Moreover, a controlled experiment is conducted to compare the operation efficiency of the VR whiteboard with two other whiteboards, including the physical whiteboard and electronic whiteboard. It shows that the VR whiteboard improves collaboration efficiency significantly compared to the mouse-based electronic whiteboard and even compatible with the traditional whiteboard.

In recent years, the aging of agricultural workers has become more serious due to the declining birthrate and aging population. Thus, fewer people are teaching young people harvesting techniques, such as how to pick a peach from the tree. So far, such training practice is proofed to be costly since its unit price is high (e.g. each one costs about two to five USD in Japan) and the peach is prone to be damaged for incorrect pick approach. Moreover, for the correct pick approach, it requires to distribute the force over the hand to avoid damaging the delicate peach. But for the traditional training method, a learner needs to repeat the trial-and-error process to get the correct hand feeling since the hand force is invisible. In this paper, to visualize the hand force, we propose a E-textile tactile glove. The glove is sewed with soft piezo-resistive sensors which can detect 26 points of force all over the hand. And a 2D GUI is designed for hand force visualization. According to the qualitative experiments, the tactile glove can detect and visualize the hand strength distribution in real time.

Image-guided and robotic surgery based on endoscopic imaging technologies can enhance cancer treatment by ideally removing all cancerous tissue and avoiding iatrogenic damage to healthy tissue. Surgeons evaluate the tumor margins at the cost of impeding surgical workflow or working with dimmed surgical illumination, since current endoscopic imaging systems cannot simultaneous and real-time color and near-infrared (NIR) fluorescence imaging under normal surgical illumination. To overcome this problem, a bio-inspired multimodal 3D endoscope combining the excellent characteristics of human eyes and compound eyes of mantis shrimp is proposed. This 3D endoscope, which achieves simultaneous and real-time imaging of three-dimensional stereoscopic, color, and NIR fluorescence, consists of three parts: a broad-band binocular optical system like as human eye, an optical relay system, and a multiband sensor inspired by the mantis shrimp’s compound eye. By introducing an optical relay system, the two sub-images after the broad-band binocular optical system can be projected onto one and the same multiband sensor. A series of experiments demonstrate that this bio-inspired multimodal 3D endoscope not only provides surgeons with real-time feedback on the location of tumor tissue and lymph nodes but also creates an immersive experience for surgeons without impeding surgical workflow. Its excellent characteristics and good scalability can promote the further development and application of image-guided and robotic surgery.

It is important for many applications to capture hand movements with high accuracy to achieve the natural human-computer interaction, such as games, robotics, rehabilitation, and virtual reality (VR). An ideal hand motion capture solution requires good mobility, unobtrusiveness, and high accuracy. In this demo, we show a hand motion capture system including inertial sensor based data gloves with the square-root cubature Kalman Filter multi-sensor fusion algorithm and a biomechanics sensor-to-segment calibration method. The absolute error of the joint angle is measured. As the result, the proposed system shows good accuracy in both static (RMSE = 1.5°) and dynamic (RMSE = 6.6°) conditions.

Power Line Communication (PLC) is a communication method that using the existing power line network as the information transmission carrier. Medium voltage (MV) underground power line can provide functions such as power quality measurement and remote control of the electrical network. However, due to the complex channel and noise model, the transmission reliability in PLC suffers a lot. As probabilistic amplitude shaping (PAS) shows a significant improvement in frame error rate (FER) in wireless channels, we aim to apply PAS in MV power line channels in this paper to evaluate its performance. First, the MV power line channel is modeled as a multiple-input multiple-output (MIMO) channel in the presence of impulsive noise and background noise. Next, we introduce the basic idea of PAS and implement it with bipolar amplitude shift keying (ASK) modulation and low-density parity-check (LDPC) codes. Finally, we compare the FER with or without PAS in the power line channel. Numerical results show that, to achieve the same FER, adopting PAS in the power line channel can save 40% transmitting power under 64-QAM modulation.

As we known, time synchronization is a crucial component of wireless sensor networks (WSNs), especially when performing the operation such as data fusion. In this paper, we revisit this problem for our proposed wireless motion capture (MoCap) nodes, namely WonderSense (WS), and present an easy and economical way to compensate time drift. We first explain the work principles of the proposed nodes. This part includes the overview of the device, and the approach to compensate the time drift for the built MoCap system. Then, the extended Kalman filter (EKF) for attitude estimation is implemented and briefly described. At last, several experiments are conducted to verify the feasibility of the time drift compensation method and validate the time drift compensated MoCap system during human walking. Results indicate that, on one hand, the proposed time drift compensation method is feasible, on the other hand, it is proved that the time drift compensated MoCap system is viable and reliable.

Zhang et al. (1) raise concerns over the characteristics and change of surface water bodies (WBs) in China reported in our article in PNAS (2). We believe these concerns are not well grounded. The dynamic nature of WBs results in a constant change in their number and extent over large land areas. Therefore, it is important to define a temporal window from whence the WB calculations are referenced, as Zhang et al. (1) highlight. Feng et al. (2) reference previous studies that relied on satellite snapshots acquired at one, or a few, points in time to study WBs across China. These studies, while useful for characterizing WB conditions at the specific points in time, could not accurately estimate the maximum extent, … [↵][1]1To whom correspondence may be addressed. Email: shuguang.liu{at}yahoo.com. [1]: #xref-corresp-1-1

We use a chaotic evolution algorithm to optimize the parameter of Gaussian kernel function in the kernel methodbased autoencoder. Kernel method-based autoencoder is an unsupervised learning algorithm with the objective of learning a representation for a set of data. Kernel methods play an important role in building a kernel method-based autoencoder. There are some options for selecting kernel functions, such as Gaussian kernel, polynomial kernel, and Laplacian kernel, etc. In each case, we are required to identify the parameters satisfying the specified requirements or problems. Unfortunately, in some cases, because of a large range of parameters, we can not select proper parameters manually. Chaotic evolution algorithm is one of the optimization algorithms, intending to obtain optimal solutions for a problem, given its certain solution search range. We take advantage of chaotic evolution algorithm to tune parameters automatically for Gaussian kernel function in this work. We found that the proposed method is an efficient and effective tool to solve the selection issue of kernel method-based autoencoder.

In remote AR conference or classroom, attendees are not being able to write freely on the whiteboard as they do in a real conference or classroom. In this paper, we solved the problem by generating an AR whiteboard in the AR space, using AR controller to write on AR whiteboard and outputting the character drawn there in a text file as a memo. The pathfinding experiment results show that the AR whiteboard can improve communication efficiency by 72% compared with the traditional real space method.

Attitude estimation is an important part for navigation of mobile robotics and unmanned aerial vehicle (UAV) control. Although the Extended Kalman Filter (EKF) can be done typically, the trend is to use Sigma-Point Kalman Filter (SPKF) instead due to its higher accuracy and robustness in harsh environment. The only drawback of such system is the higher computation cost. In order to accelerate the system, most approaches based on Field Programmable Gate Arrays (FPGA) are proposed in the past but too specific, which is not reusable and the high price for design complexity. With looking for re-usability, we present an IP core called matrix operation accelerator in this paper. Moreover, we do the verification on Zynq-7020, the experimental result shows that the proposed scheme can reduce about 50% computing time and save silicon as well.

Malware issue puzzles software users. Under the development of Internet, software technology has been involved in all parts of life such as Robot Control Software. There have been many methods to detect malicious Android applications. However, most of these are single behavior analysis models. In order to better detect the malicious behavior for more fields, we propose a malicious behavior analysis model based on system call and function interface with machine learning algorithm training. To clear the analysis model, we discuss the reasonable feature selection, data collection and analyze the implementation of static methods, dynamic methods and hybrid methods. In addition, we compare several important algorithm models to get the optimal model. Experimental result shows that our proposed optimal model has a higher accuracy than traditional way.

The blurred synthesis evaluation arithmetic based on analytic network process and systemic security evaluation target system is designed, and the main technology and system realization of cyberspace security evaluation system are presented, which is realized based on fight theory’s attack and defense affection evaluation and systemic security evaluation technology. The technology and system can provide a systemic security evaluation to information system target’s security and put forward security resolve scheme suggestions, which have important effect and meaning on accelerating China’s cyberspace war ability.

CVSS is one of the most representative quantitative assessment algorithms for security vulnerabilities. The calculated values of vulnerability harm according to the given index and formula. However, there is still insufficient objectivity of CVSS. Expert System is a kind of expert system for vulnerabilities; the severity of vulnerability can be assessed by experts based on the Expert System. The objectivity of Expert System and CVSS is analyzed, and CVSS is revised based on Expert System. The relationship between the above systems and the CWE, and between above systems and Product is analyzed, respectively, in the process. A new way of using Expert System is put forward based on CWE, namely, CWE Cycle Sorting Algorithm, in order to sort the average of vulnerability severity. Meanwhile, CWE Sort Factor is put forward based on CWE Cycle Sorting Algorithm to modify CVSS. The result is closer to the Expert System in terms of objectivity.

Active power decoupling (APD) techniques prove to be efficient in cutting passive component size for single-phase AC/DC converter, usually against bulky electrolytic capacitors (E-Cap) on DC side. Generally, no adding extra switches is expected to realize E-Capless function as well as original power factor correction (PFC). This paper proposes a multiplexing modulation strategy with optimized power-decoupling reference algorithm, employing a small LC branch as power decoupling circuit to replace large electrolytic capacitors, which could further reduce dc-link capacitance. Both simulation and experimental results for a lkW prototype are presented in this paper to validate the performance of proposed method.

Active power decoupling (APD) techniques are widely used in single-phase PWM rectifiers under 10 kilowatts, which prove to be efficient in cutting passive component size (usually against bulky electrolytic capacitors at dc side). However, it is found that the reactive power caused by boost inductor cannot be ignored especially in low input voltage, high power situation with large input inductance. Such reactive power could bring an obvious phase-shifted angle ϕ to dc side current, which contributes to the steady state error in decoupling reference for APD circuit. This paper presents a strategy by modifying ϕ to achieve the ac-side inductor reactive power compensation (IRPC) without adding any hardware, which could further reduce dc-link capacitance. Compared with the traditional control method that usually neglects reactive power stored by inductor, a more effective power-decoupling IRPC scheme is validated experimentally with a 1kW prototype.

Vehicle to grid (V2G) technology uses electric vehicles (EVs) as energy storage devices for the power grid along with peak cutting and grain filling function. The bidirectional converter plays an essential part in V2G technology since it allows energy to flow in both directions based on the grid dispatching signal. However, in two-stage bidirectional on-board-charge (OBC) with CLLC circuit, traditional frequency conversion control can not achieve smooth mode-switching, so that the OBC can not response to the power grid dispatching signal quickly. In this paper, a control strategy combining frequency conversion control and phase-shifting control is proposed, which can realize smooth mode-switching of CLLC circuit. The control strategy is validated by a 3kW two-stage OBC prototype integrating Totem-pole and CLLC circuit.

Distributed Permutation Flowshop Scheduling Problem (DPFSP) has become a hot issue in recent years. However, DPFSP with uncertain processing times (DPFSP_UPT) has not been addressed so far although real manufacturing systems often encounter various uncertain factors which make processing times uncertain. In this paper, two local-search based metaheuristics, i.e., an Iterated Greedy algorithm (sIG) and an Iterated Local Search algorithm (sILS), are presented to solve the DPFSP_UPT with makespan criterion. A robust model of expect-risk rule is used to describe the DPFSP_UPT. A valid heuristic based on the well-known NEH heuristic is proposed to initialize sIG and sILS. An acceleration algorithm is adapted to save computational efforts. A destruction procedure with dynamic sizes is presented to enhance the exploration capability for sIG. A feature with the characteristic that small changes in the solution can preserve the properties of good solutions in the iterative process is applied by sILS. Both sIG and sILS use multiple local search methods to produce promising solutions by exploiting diverse search areas. Extensive experiments show that the proposed sIG and sILS perform significantly better than the five competing algorithms adapted in the literature, but there is no significant difference between sIG and sILS and each has its own advantages for different instances.

Knowledge sharing plays an important role in promoting innovation and helping improve R&D team performance in the digital age. Based on the evolutionary game theory, this study develops an evolutionary game model of knowledge sharing in R&D teams in order to explore its system evolution path, the evolutionary stability strategy, and the influencing mechanism in knowledge sharing. Then using a simulation model, this study examines the dynamic evolution process of knowledge sharing within R&D teams. The results show that the effectiveness of knowledge sharing in the R&D teams can be promoted by R&D team members’ cognitive ability, knowledge absorption ability, knowledge transformation ability, knowledge innovation ability, and the degree of knowledge complementarity within teams. The simulation results further show that reducing the environmental risk can also effectively improve R&D teams’ innovation performance. The findings of this study thus provide evidence for knowledge sharing as an important route to sustainable development.

A novel method of geometric dimension assisted absolute phase recovery in 3D shape measurement is presented. This method mainly includes two steps: (a) using the window Fourier filter-quality guided phase unwrapping algorithm to obtain the relative phase distribution of the tested object; (b) using the geometric dimension of the object as a clue to convert the relative phase distribution to the absolute phase distribution. This method is convenient, and only three fringe image acquisitions are needed to recover the absolute phase. In addition, there is no limit to the depth range of the tested object. The correctness of this method is verified by several experiments.

The spatialization of population is the process of rasterizing demographic data, which is of great significance for government decision-making and promoting the coordinated and sustainable development of social economy, resources and environment. In the continuous research of multiple linear regression, spatial interpolation, the use of multi-source heterogeneous data, random forest and other methods, a batch of highly representative global and national-scale population spatial databases have been born. Therefore, through the super-resolution of deep learning the method is more valuable for population low-scale sampling research and end-to-end mapping to achieve high-precision downscaling and enhance the robustness and generalization of the training model. This paper is based on the population statistics of the counties in Chongqing municipality in 2010, through the area-weighted rasterization coupled sub-pixel convolutional neural network (ESPCN) to learn the global and local features of the population to achieve 500m resolution population spatialization, and compare it with the other 2 compare plan. The results show that the training model using this method has the lowest mean square error, the smallest residual interval, and the best picture effect.

Distributed Permutation Flowshop Scheduling Problem (DPFSP) has become a research hotspot in recent years. However, as a service level objective, the Total Weighted Earliness and Tardiness (TWET) has not been addressed so far. Due to the importance of the service level objective in modern industry, we deal with the minimization of the TWET for the DPFSP with due windows . An Iterated Greedy (IG) algorithm, namely IG with Idle Time insertion Evaluation (IG I T E ), is proposed. In the algorithm, an adapted NEH heuristic with five rules based on the unit earliness weight and unit tardiness weight, the due date, and the smallest slack on the last machine is used to generate an initial solution. Destruction procedure with a dynamic size is provided to enhance the exploration capability of the algorithm. Idle time insertion method is utilized to make the completion time of jobs within the due windows or as close to the due windows as possible. A large number of experiments show that the presented algorithm performs significantly better than the five competing algorithms adapted in the literature. The performance analysis shows that the IG I T E is the most appropriate for the DPFSP with due windows among the tested algorithms.

Taking the passenger aircraft A320 as the research object, the trajectory optimization of the aircraft climbing section under wind disturbance is carried out. Through the analysis of the aircraft’s descent process, the kinematics model, fuel consumption model and atmospheric model of the aircraft are established, and the objective function and constraints of the problem are obtained. The quantum genetic algorithm is used to transform the optimal control problem into a nonlinear programming problem. The optimal trajectory of the aircraft’s shortest descent time is obtained by simulation. The effects of different cost indices on the fall time and fuel consumption are discussed.

This work presents an investigation of multipath interference (MPI) of operator's existing PON network. The reflectance of connectors within the ODN link are measured, and the MPI penalty of the link is calculated for some candidate modulation schemes. The result will be great helpful to the ongoing development next-generation FTTx standards.

We investigate the characteristics of hard margin support vector machine using a real-world handwriting data. The number of support vectors and distance of margin are two investigated objectives in this work. The structural risk of support vector machine is influenced by the distance of margin that is decided by settings of the kernel function and it parameter. We found that number of support vectors are as the same when the parameter of kernel functions are changed, but the distance of margin is highly influenced by the parameter of kernel functions, which is related to the structural risk of established support vector machine. The handwriting data are not always mixed together in their original representation space. Some of them are linear separable, but some of them are not. We should design a group of classifiers to handle this real-world handwriting application by considering these characteristics and the primary discoveries of this work.

In recent years, most countries around the world have faced increasing pressures in the realm of emergency management than ever before. Medical service organization selection is one of the most vital facets of emergency management. Meanwhile, during the selection process, many criteria may conflict with one another and information is uncertain, rendering decision-making processes complex. Hence, multi-objective optimization, fuzzy way and stochastic theories serve as suitable means of addressing such problems. In this paper, a fuzzy multi-objective linear model is developed to overcome medical service organization selection issues and uncertain information. Meanwhile, a fuzzy objective and weight are applied to enable the decision-maker to select suitable schemes while considering stochastic medical service demand. Moreover, real data cannot been obtained. Hence, according to actual conditions, we assume relative information. For illustrative purposes, a numerical example is presented to verify the effectiveness of the proposed model from experimental data.

Term translation of Chinese historical classics is very difficult and time-consuming work, and using term alignment methods to extract term translation pairs is of great help for historical term translation. However, the limited bilingual corpora resources of historical classics and special morphology of the ancient Chinese result in poor performance of term alignment. To this end, this paper proposes a historical term alignment method using modern Chinese as a pivot language. The method first identifies English terms by rules, then aligns them from English to modern Chinese and then from modern Chinese to ancient Chinese. The use of English-modern Chinese corpus and modern-ancient Chinese corpus instead of English-ancient Chinese corpus solves the shortage problem of the parallel corpus. Moreover, using modern Chinese as a pivot language effectively reduces the alignment errors caused by the abbreviations and the interchangeable characters of ancient Chinese. In the term alignment experiment on Shiji, our method outperformed the direct alignment method significantly, which proves the validity of our method.

With the development of smart grid and new energy technology, V2G (vehicle-to-grid) has become a new research hotspot, and bidirectional converter plays an important role in it. In this paper, a bidirectional AC/DC converter using SiC device is developed, and unipolar PWM modulation is adopted in consideration of efficiency. Compared with other modulation methods that are widely used, unipolar modulation has the advantages of low switching loss, small current ripple and small common mode noise, however the current is distorted at zero-crossing. With the investigation on the characteristics of the current zero-crossing distortion based on the AC/DC converter prototype, an optimized control strategy which generates the switching signal for the line frequency bridge based on the control loop output reference (CLOR) instead of the grid voltage is proposed. The performance of the analyses and the proposed strategy is verified by experiments.

Mobile application for campus is an integral part of the construction of wisdom campus. CampusService - a campus information service system is designed, and is comprised of three parts, Android client, Web client and server, which are all designed according to MVC pattern. C/S architecture is adopted between the Android client and the server, B/S architecture is adopted between the Web client and the server, and JSON is used as a unified data interface for accomplishing the interactions. The architectures of three parts are described, and key technologies are represented, including database access optimization, Android display, transaction management and authentication, among which database access optimization technology has two respects: using C3P0 to maintain connection pool and prefetching data based on their popularity. According to testing, the system implements functions of user management, intramural announcements, interaction with friends, activity participation, and campus maps, and it has good performance. The system provides an efficient platform for real-time communication among teachers, students and school authority.

Detection and analysis of activities of daily living (ADLs) are important in activity tracking, security monitoring, and life support in elderly healthcare. Recently, many research projects have employed wearable devices to detect and analyze ADLs. However, most wearable devices obstruct natural movement of the body, and the analysis of activities lacks adequate consideration of various real attributes. To tackle these issues, we proposed a two-fold solution. First, regarding unobtrusive detection of ADLs, only one small device is worn on a finger to sense and collect activity information, and identifiable features are extracted from the finger-related signals to identify various activities. Second, to reflect realistic life situations, a weighted sequence alignment approach is proposed to analyze an activity sequence detected by the device, as well as attributes of each activity in the sequence. The system is validated using 10 daily activities and 3 activity sequences. Results show 96.8 % accuracy in recognizing activities and the effectiveness of sequence analysis.

In this paper, a new handwriting recognition system is proposed with an inertial sensor attached on the fingertip to automatically digitalize the cursive handwriting characters. Specifically, it focus on the issues and methods about the sensor node design, automatic spotting, reconstruction of writing trajectory, and feature selection. The performance of the system was evaluated on a set of 36 vocabularies (10 digits and 26 English lower-case characters). The recognition rate is 88.0% on SVM with user-independent 5 fold cross evaluation protocol.

Human Activity Recognition (HAR) is a critical technology for seniors home care. In this paper, we present the system implementation and experimental study on detection of both of the daily activities and accidental event (Fall Down) with multiple inertial sensors on body. The overall accuracy are 94.3% on recognition of 10 kinds of daily activities with kNN in user-independent evaluation. Moreover, the experiment shows that the combination of multiple sensors on different locations of upper, middle, lower body parts can improve both of the accuracy and stability (one node: 78.1%±8.0%, two nodes: 90.8%±4.7%, and three nodes are 94.3%±4.4%). Finally, we investigate the detection of fall down from the other ten daily activities. The accuracy is 87.5% with mean and standard deviation as the features, and improved to 100% with energy as the additional feature.

In this research, we present a practical method to detect the insufficient brushing points. The system consists of a sensor node and a visual interface. The sensor node can be clipped onto the handler of a toothbrush while brushing the teeth and the brushed regions will be realtime displayed on the visual interface. A six-axis inertial sensor inside the node is used to detect the scrubbed areas, so that the user can realize the insufficient scrubbed areas in an intuitive way. All teeth are divided into 16 regions to evaluate the detection accuracy. The user dependent accuracy is 88%, and user independent accuracy is 72% across five users. Moreover, Two users took part in a learning effect experiment. After one-week usage, their insufficient toothbrushing points reduced by 80% and 77% perspectively.

Information distribution regarding immediately correlated contents is a big challenge in local communities, such as a university campus. One of the main reasons is that a campus' local network is uncovered by public search engines, and most of the local Web sites in a community operate like information silos. A secondary reason is that the daily activities in a local community are highly dynamic, people might be annoyed by the non-correlated information or they just ignore all of it, regardless of importance. In this paper, the proposed campus digital signage aims at connecting local users to correlated information as the user passes by. To this end, we build the server side with two components, one is the information collector that will automatically pull the campus information from other sites, and the other is the information poster to allow qualified users to post content. We also design a wearable client side with a BLE (Bluetooth low energy) proximity sensor to recognize approaching user's identity and dynamically push their correlated information to the digital signage display.

Online segmentation of continuous gestures is a key issue in accelerometer-based gesture recognition. The challenge is to segment short-duration, ambiguous, and individually different gestures by an accurate method for online applications. In this paper, we propose a threshold-based method for the online segmentation of one-stroke gestures using a single tri-axial accelerometer on a finger. The method first obtains a proper initial threshold based on Bayes decision theory to ensure accurate initial segmentation, and then an adaptation mechanism is designed to automatically adjust the threshold to improve user-dependent accuracy. A dataset of over 1,200 samples from 12 gestures and 25 subjects is created for evaluation. The results show that our method achieves high segmentation precision/recall and high user-dependent adaptability with low computational complexity.

Hidden Markov Model (HMM)-based recognition methods are very commonly used for some applications and can be highly accurate. However, they have a high computational complexity that creates problems when they are used for gesture recognition on resource-constrained wearable devices. In this paper, we propose a pre-classification method to reduce recognition complexity by dividing gesture vocabularies into groups, and maintain, even improve, the recognition accuracy by adaptively adjusting the HMMs for different groups. The technique consists of three tasks: gesture grouping, group modeling, and gesture modeling. Gesture grouping is performed using a K-means++ algorithm; the groups are modeled using a table-based method; and the gestures are modeled using an HMM-based approach. We evaluated the pre-classification method using real data collected by a tiny finger-worn device called a Magic Ring. The complexity of our method is much less than the standard Hidden Markov Model, without any loss of accuracy.

Cloud computing advocates a promising paradigm that facilitates the access within heterogeneous services, platforms, and end users. However, platforms (or host servers) have confined to devices which require a considerable computing resources. In this case, solutions concerning the efficient use of pervasive devices with constrained resources become an open issue. This study investigates the seamless connection between embedded devices and cloud resources to enhance the capability of computing and furthermore provide context-aware services. A method for wireless program dissemination and boot loading is proposed to transfer necessary information and resources between service and target device(s). The experiment results on time delay and energy cost demonstrate the feasibility and performance.

This paper proposes a symbol flipping decoding algorithm based on prediction for non-binary LDPC codes, considering not only soft reliability, but also hard reliability. It provides 2.6 dB improvement compared with the weighted Algorithm B.

Recently, gesture recognition by the motion sensor embedded in wearable device has gradually been a popular research topic in pervasive computing. Traditionally, most of gesture recognition researches are based on sensor waveform of time and frequency domains. Another method based on trajectory estimation is emerging in gesture recognition. This method seems like more effect since it can extract more features compared with waveform-based recognition. But a quantitative experiment have not been taken to compare the trajectory-based method with waveform-based method. In this paper, we present an argument that the method based on trajectory estimation can provide higher and more stable recognition. To verify our viewpoint, we performed a controlled experiment based on an online finger gesture recognition system. As the result of our experiment, comparing trajectory estimation method with waveform method, we found that the recognition rate of trajectory-based method can reach 96.20% (SD) in userdependent and 91.20% accuracy in user-independent experiments respectively, which outperform waveform method by 4.9% and 3.9%.