Earticle

Plantar pressure data can be used not only for walking patterns in daily life, but also for eating, health care, and disease prevention. For this reason, the importance of plantar pressure measurement has recently increased. However, most systems that can measure both static and dynamic plantar pressure at the same time are expensive, not portable, and not universal. In this study, we propose a system that effectively reduces the number of sensors in plantar pressure system. Through this, we want to increase the economics and practicality by reducing the size and weight of the system, as well as the power consumption. First, for static plantar pressure and dynamic plantar pressure, the values measured by existing precision instruments are analyzed to determine how many measurement parts the insole is divided into. Next, for the divided measuring parts, the position of the sensor is determined by calculating the Center of Pressure (COP) for each part with the values of all dynamic and static plantar pressure sensors. Finally, in order to construct a personalized plantar pressure measurement system, we propose a weighting method for the static plantar pressure COP and the dynamic plantar pressure COP for each part.

The public data portal provides various public data created by the government in the form of files and open APIs. In order to increase the usability of public open data, a variety of information should be provided to users and should be convenient to use for users. This requires the structured data design plan of the public data. In this paper, we propose a data design method to improve the usability of the Seoul subway public data. For the study, we first identify some properties of the current subway public data and then classify the data based on these properties. The properties used as classification criteria are stored properties, derived properties, static properties, and dynamic properties. We also analyze the limitations of current data for each property. Based on this analysis, we classify currently used subway public data into code entities, base entities, and history entities and present the improved design of entities according to this classification. In addition, we propose data retrieval functions to increase the utilization of the data. If the data is designed according to the proposed design of this paper, it will be possible to solve the problem of duplication and inconsistency of the data currently used and to implement more structural data. As a result, it can provide more functions for users, which is the basis for increasing usability of subway public data.

This paper proposes a new voice activity detection (VAD) method which is based on SNR and non-intrusive speech intelligibility estimation. In the conventional SNR-based VAD methods, voice activity probability is obtained by estimating frame-wise SNR at each spectral component. However these methods lack performance in various noisy environments. We devise a hybrid VAD method that uses non-intrusive speech intelligibility estimation as well as SNR estimation, where the speech intelligibility score is estimated based on deep neural network. In order to train model parameters of deep neural network, we use MFCC vector and the intrusive speech intelligibility score, STOI (Short-Time Objective Intelligent Measure), as input and output, respectively. We developed speech presence measure to classify each noisy frame as voice or non-voice by calculating the weighted average of the estimated STOI value and the conventional SNR-based VAD value at each frame. Experimental results show that the proposed method has better performance than the conventional VAD method in various noisy environments, especially when the SNR is very low.

Unsupervised neural networks have not caught enough attention until Generative Adversarial Network (GAN) was proposed. By using both the generator and discriminator networks, GAN can extract the main characteristic of the original dataset and produce new data with similar latent statistics. However, researchers understand fully that training GAN is not easy because of its unstable condition. The discriminator usually performs too good when helping the generator to learn statistics of the training datasets. Thus, the generated data is not compelling. Various research have focused on how to improve the stability and classification accuracy of GAN. However, few studies delve into how to improve the training efficiency and to save training time. In this paper, we propose a novel optimizer, named FAST-ADAM, which integrates the Lookahead to ADAM optimizer to train the generator of a semi-supervised generative adversarial network (SSGAN). We experiment to assess the feasibility and performance of our optimizer using Canadian Institute For Advanced Research – 10 (CIFAR-10) benchmark dataset. From the experiment results, we show that FAST-ADAM can help the generator to reach convergence faster than the original ADAM while maintaining comparable training accuracy results.

In this paper, we explore the details of three classic data augmentation methods and two generative model based oversampling methods. The three classic data augmentation methods are random sampling (RANDOM), Synthetic Minority Over-sampling Technique (SMOTE), and Adaptive Synthetic Sampling (ADASYN). The two generative model based oversampling methods are Conditional Generative Adversarial Network (CGAN) and Wasserstein Generative Adversarial Network (WGAN). In imbalanced data, the whole instances are divided into majority class and minority class, where majority class occupies most of the instances in the training set and minority class only includes a few instances. Generative models have their own advantages when they are used to generate more plausible samples referring to the distribution of the minority class. We also adopt CGAN to compare the data augmentation performance with other methods. The experimental results show that WGAN-based oversampling technique is more stable than other approaches (RANDOM, SMOTE, ADASYN and CGAN) even with the very limited training datasets. However, when the imbalanced ratio is too small, generative model based approaches cannot achieve satisfying performance than the conventional data augmentation techniques. These results suggest us one of future research directions.

We design a reconfigurable optical node for metropolitan WDM networks, and numerically study the capability of the node in the optical signal level. Unlike a long-haul WDM system, major limitations of metropolitan WDM systems are power loss, fiber dispersion and optical signal-to-noise ratio(OSNR) degradation due to EDFAs. Therefore, we include the behaviors of transmitter and receiver, and fiber, EDFAs, and optical filters(MUX/DeMux) in numerical simulations with varying parameters over wide range. From simulation results, we can identify the maximum span numbers for OC-48 and OC-192 to achieve BER<10-12 using the node structure at various received powers and residual dispersions.

3D scanners are needed in various fields, and their usage range is greatly expanded. In particular, it is being used to reduce costs at various stages during product development and production. Now, the importance of quality inspection in the manufacturing industry is increasing. Structured optical system applied in this study is suitable for measuring high precision of mold, press work, precision products, etc. and economical and effective 3D scanning system for measuring inspection in manufacturing industry can be implemented. We developed Structured light 3D scanner which can measure high precision by using Digital Light Processing (DLP) projector and camera. In this paper, 3D image scanner based on structured optical system can realize 3D scanning system economically and effectively when measuring inspection in the manufacturing industry.

The purpose of this research is to examine the relationship between corporate social responsibility (CSR), green supply chain management (GSCM) practices, and business performances. After reviewing the extensive literature, we developed a research model including five constructs: CSR, GSCM practices, environmental, economic and operational performances. We conducted the statistical analyses based on the primary data collected from a survey questionnaire, responded by 93 different company managers in the Republic of Guatemala. Furthermore, we utilized structural equation modeling to analyze the data and to test the hypotheses. The results of the analyses showed that there is a significant influence of CSR on the adoption of GSCM practices. It was also found that GSCM practices have a significant influence on environmental, economic and operational performances. In addition, environmental performance has a significant impact on economic and operational performance. Finally, GSCM has a mediating role on the relationship between CSR and environmental and economic performance, but not with operational performance.

We are looking the impact associated with globalization in favor of small and medium enterprises (SMEs) growth and how helped to reduce the obstacle facing SMEs growth. We used empirical analysis in order to examine the relationship underlying the globalization and its impact to SME growth. We employed primarily data from Tanzania SMEs. Further we seeks to explain the negative notion that has been created that globalization is not friendly to SME growths. We employed primary data from Tanzania SMEs. The partial least squares (PLS) was used for analysis. The conclusion has indicated that globalization has a relationship with SMEs growth and has contributed to the reduction of obstacles that inhibit SMEs growth. However study confirmed controversial result on part of availability of managers and manpower with global perspectives to influence SMEs growth. The test accepted that globalization has influenced availability of managers with global perspectives but reject the availability of these managers influences the SMEs growth The results give a clear outlook to help policy maker in policy review process, formulate base for extensive study on issues for manager perspectives and draw intervention.

Sleep apnea is a disease that causes various complications, and the polysomnography is expensive and difficult to measure. The purpose of this study is to develop an unrestricted wearable monitoring system so that patients can be examined in a familiar environment. We used a method to detect sleep apnea events and to determine sleep satisfaction by non-constrained method using SpO2 measurement sensor and 3-axis acceleration sensor. Heart rate and SpO2 were measured at the finger using max30100. After acquiring the SpO2 data of the user in real time, the apnea measurement algorithm was used to transmit the number of apnea events of the user to the mobile phone using Bluetooth (HC-06) on the wrist. Using the three-axis acceleration sensor (mpu6050) attached to the upper body, the number of times of tossing and turning during sleep was measured. Based on this data, this algorithm evaluates the patient's tossing and turning during sleep and transmits the data to the mobile phone via Bluetooth. The power source used 9 volts battery to operate Arduino UNO and sensors for portability and stability, and the data received from each sensor can be used to check the various degree between sleep apnea and sleep tossing and turning on the mobile phone. Through this study, we have developed a wearable sleep apnea measurement system that can be easily used at home for the problem of low sleep efficiency of sleep apnea patients.

We introduce a vision-based object detection method for real-time video surveillance system in low-end edge computing environments. Recently, the accuracy of object detection has been improved due to the performance of approaches based on deep learning algorithm such as Region Convolutional Neural Network(R-CNN) which has two stage for inferencing. On the other hand, one stage detection algorithms such as single-shot detection (SSD) and you only look once (YOLO) have been developed at the expense of some accuracy and can be used for real-time systems. However, high-performance hardware such as General-Purpose computing on Graphics Processing Unit(GPGPU) is required to still achieve excellent object detection performance and speed. To address hardware requirement that is burdensome to low-end edge computing environments, We propose subframe analysis method for the object detection. In specific, We divide a whole image frame into smaller ones then inference them on Convolutional Neural Network (CNN) based image detection network, which is much faster than conventional network designed for full frame image. We reduced its computational requirement significantly without losing throughput and object detection accuracy with the proposed method.