• Proceedings of the Korean Vacuum Society Conference
• /
• 2013.08a
• /
• pp.75.2-75.2
• /
• 2013

Atomic layer deposition (ALD), utilizing self-limiting surface reactions, could offer promising perspectives for future efficient energy conversion devices. The capabilities of ALD for surface/interface modification and construction of novel architectures with sub-nanometer precision and exceptional conformality over high aspect ratio make it more valuable than any other deposition methods in nanoscale science and technology. In the context, a variety of researches on fabrication of active materials for energy conversion applications by ALD are emerging. Among those materials, one-dimensional nanotubular titanium dioxide, providing not only high specific surface area but also efficient carrier transport pathway, is a class of the most intensively explored materials for energy conversion systems, such as photovoltaic cells and photo/electrochemical devices. The monodisperse, stoichiometric, anatase, TiO2 nanotubes with smooth surface morphology and controlled wall thickness were fabricated via low-temperature template-directed ALD followed by subsequent annealing. The ALD-grown, anatase, TiO2 nanotubes in alumina template show unusual crystal growth behavior which allows to form remarkably large grains along axial direction over certain wall thickness. We also fabricated dye-sensitized solar cells (DSCs) introducing our anatase TiO2 nanotubes as photoanodes, and studied the effect of blocking layer, TiO2 thin films formed by ALD, on overall device efficiency. The photon convertsion efficiency ~7% were measured for our TiO2 nanotubebased DSCs with blocking layers, which is ~1% higher than ones without blocking layer. We also performed open circuit voltage decay measurement to estimate recombination rate in our cells, which is 3 times longer than conventional nanoparticulate photoanodes. The high efficiency of our ALD-grown, anatase, TiO2 nanotube-based DSCs may be attributed to both enhanced charge transport property of our TiO2 nanotubes photoanode and the suppression of recombination at the interface between transparent conducting electrode and iodine electrolytes by blocking layer.

• Phonetics and Speech Sciences
• /
• v.11 no.2
• /
• pp.57-64
• /
• 2019

In this paper, we construct a Korean text-to-speech system using the Merlin toolkit which is an open source system for speech synthesis. In the text-to-speech system, the HMM-based statistical parametric speech synthesis method is widely used, but it is known that the quality of synthesized speech is degraded due to limitations of the acoustic modeling scheme that includes context factors. In this paper, we propose an acoustic modeling architecture that uses deep neural network technique, which shows excellent performance in various fields. Fully connected deep feedforward neural network (DNN), recurrent neural network (RNN), gated recurrent unit (GRU), long short-term memory (LSTM), bidirectional LSTM (BLSTM) are included in the architecture. Experimental results have shown that the performance is improved by including sequence modeling in the architecture, and the architecture with LSTM or BLSTM shows the best performance. It has been also found that inclusion of delta and delta-delta components in the acoustic feature parameters is advantageous for performance improvement.

• International journal of advanced smart convergence
• /
• v.10 no.1
• /
• pp.1-11
• /
• 2021

Document page segmentation is an important step in building a quality optical character recognition module. The study examined already existing work on the topic of page segmentation and focused on the development of a segmentation model that has greater functional significance for application in an organization, as well as broad capabilities for managing the quality of the model. The main problems of document segmentation were highlighted, which include a complex background of intersecting objects. As classes for detection, not only classic text, table and figure were selected, but also additional types, such as signature, logo and table without borders (or with partially missing borders). This made it possible to pose a non-trivial task of detecting non-standard document elements. The authors compared existing neural network architectures for object detection based on published research data. The most suitable architecture was RetinaNet. To ensure the possibility of quality control of the model, a method based on neural network modeling using the RetinaNet architecture is proposed. During the study, several models were built, the quality of which was assessed on the test sample using the Mean average Precision metric. The best result among the constructed algorithms was shown by a model that includes four neural networks: the focus of the first neural network on detecting tables and tables without borders, the second - seals and signatures, the third - pictures and logos, and the fourth - text. As a result of the analysis, it was revealed that the approach based on four neural networks showed the best results in accordance with the objectives of the study on the test sample in the context of most classes of detection. The method proposed in the article can be used to recognize other objects. A promising direction in which the analysis can be continued is the segmentation of tables; the areas of the table that differ in function will act as classes: heading, cell with a name, cell with data, empty cell.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.19 no.4
• /
• pp.701-708
• /
• 2024

The advent of transformer architectures in deep learning has been a major breakthrough in natural language processing research. Object name recognition is a branch of natural language processing and is an important research area for tasks such as information retrieval. It is also important in the biomedical field, but the lack of Korean biomedical corpora for training has limited the development of Korean clinical research using AI. In this study, we built a new biomedical corpus for Korean biomedical entity name recognition and selected language models pre-trained on a large Korean corpus for transfer learning. We compared the name recognition performance of the selected language models by F1-score and the recognition rate by tag, and analyzed the errors. In terms of recognition performance, KlueRoBERTa showed relatively good performance. The error analysis of the tagging process shows that the recognition performance of Disease is excellent, but Body and Treatment are relatively low. This is due to over-segmentation and under-segmentation that fails to properly categorize entity names based on context, and it will be necessary to build a more precise morphological analyzer and a rich lexicon to compensate for the incorrect tagging.

• Proceedings of the Korean Society for Noise and Vibration Engineering Conference
• /
• 2010.05a
• /
• pp.47-47
• /
• 2010

Over the past 20 years, constant progress in noise and vibration (NVH) engineering has enabled to constantly advance quality and comfort of operation and use of really any products - from automobiles to aircraft, to all kinds of industrial vehicles and machines - to the extend that for many products, supreme NVH performance has becomes part of its brand image in the market. At the same time, the product innovation agenda in the automotive, aircraft and really many other industries, has been extended very much in recent years by meeting ever more strict environmental regulations. Like in the automotive industry, the drive towards meeting emission and CO2 targets leads to very much accelerated adoption of new powertrain concepts (downsizing of ICE, hybrid-electrical...), and to new vehicle architectures and the application of new materials to reduce weight, which bring new challenges for not only maintaining but further improving NVH performance. This drives for innovation in NVH engineering, so as to succeed in meeting a product brand performance for NVH, while as the same time satisfying eco-constraints. Product innovation has also become increasingly dependent on the adoption of electronics and software, which drives for new solutions for NVH engineering that can be applied for NVH performance optimization of mechatronic products. Finally, relentless pressure to shorten time to market while maintaining overall product quality and reliability, mandates that the practice and solutions for NVH engineering can be optimally applied in all phases of product development. The presentation will first review the afore trends for product and process innovation, and discuss the challenges they represent for NVH engineering. Next, the presentation discusses new solutions for NVH engineering of products, so as to meet target brand values, while at the same time meeting ever more strict eco constraints, and this within a context of increasing adoption of electronics and controls to drive product innovation. NVH being very much defined by system level performance, these solutions implement the approach of "Model Based System Engineering" to increase the impact of system level analysis for NVH in all phases of product development: - At the Concept Phase, to be able to do business case analysis of new product concepts; to arrive at an optimized and robust product architecture (e.g. to hybrid powertrain lay-out, to optimize fuel economy); to enable target cascading, to subsystem and component level. - In Development Phase, to increase realism and productivity of simulation, so as to frontload virtual validation of components and subsystems and to further reduce reliance on physical testing. - During the final System Testing Phase, to enable subsystem testing by a combination of physical testing and simulation: using simulation models to simulate the final integration context when testing a subsystem, enabling to frontload subsystem testing before final system integration is possible. - To interconnect Mechanical, Electronical and Controls engineering, in all phases of development, by supporting model driven controls engineering (MIL, SIL, HIL). Finally, the presentation reviews examples of how LMS is implementing such new applications for NVH engineering with lead customers in Europe, Asia and US, with demonstrated benefits both in terms of shortening development cycles, and/or enabling a simulation based approach to reduce reliance on physical testing.

• Journal of KIISE:Computer Systems and Theory
• /
• v.29 no.7
• /
• pp.422-430
• /
• 2002

For a time-multiplexed FPGA, a circuit is partitioned into several subcircuits, so that they temporally share the same physical FPGA device by hardware reconfiguration. In these architectures, all the hardware reconfiguration information called contexts are generated and downloaded into the chip, and then the pre-scheduled context switches occur properly and timely. Since the maximum number of the LUT's required in the same time determines the size of the chip used in the synthesis, it needs to be minimized, if possible. Many previous work use their own approaches, which are very similar to either scheduling method in high level synthesis or multi-way circuit partitioning method, to solve the problem. In this paper, we propose a method which estimates the lower bound on the number of LUT's without performing any actual synthesis. The estimated lower bounds help to evaluate the results of the previous work. If the estimated lower bound on the number of LUT's exactly matches the number of LUT's of the result from the previous work, the result must be optimal. In contrast, if they do not match, the following two cases are expected : the more exact lower bound may exist, or we might find the new synthesis result better than the result from the previous work. Experimental results show that our lower bound estimation method is very accurate. In almost al] cases experimented, the estimated lower bounds on the number of LUT's exactly match those of the previous synthesis results respectively, implying that the best results from the previous work are optimal as well as our method predicted the exact lower bound for those examples.

• Journal of KIISE:Computer Systems and Theory
• /
• v.31 no.11
• /
• pp.626-634
• /
• 2004

In a client-server model, network server systems suffer from both heavy communication and computational loads. While communication channels become increasingly speedy, the existing protocol stack architectures still include mainly three performance bottlenecks of protocol stack processing, system call, and network interrupt overheads. To address these obstacles, in this paper we present a host-independent network system where a network interface card (NIC) is utilized in an efficient manner. First, by offloading network-related portion to the NIC, the host can fully utilize its processing power for other useful purposes. Second, it eliminates the system call overhead, such as context-switching and memory copy operations, since the host communicates with the NIC through its user-level libraries. Third, it a] so reduces the network interrupt operation count as the host handles the interrupt in a segment instead of a packet. The experimental results show that the proposed network system reduces the host CPU overhead for communication system by 68-71%. It also shows that the proposed system improves the communication speed by 11-83% under heavy computational and communication load conditions.

• Journal of the Korea Institute of Information Security & Cryptology
• /
• v.17 no.4
• /
• pp.115-129
• /
• 2007

The ubiquitous and autonomic computing environments is open and dynamic providing the universal wireless access through seamless integration of software and system architectures. The ubiquitous computing have to offer the user-centric pervasive services according to the wireless access. Therefore the roaming services with the predefined security associations among all of the mobile devices in various networks is especially complex and difficult. Furthermore, there has been little study of security coordination for realistic autonomic system capable of authenticating users with different kinds of user interfaces, efficient context modeling with user profiles on Smart Cards, and providing pervasive access service by setting roaming agreements with a variety of wireless network operators. This paper proposes a Roaming Coordinator-based security management framework that supports the capability of interoperator roaming with the pervasive security services among the push service based network domains. Compared to traditional mobile systems in which a Universal Subscriber Identity Module(USIM) is dedicated to one service domain only, our proposed system with Roaming Coordinator is more open, secure, and easy to update for security services throughout the different network domains such as public wireless local area networks(PWLANs), 3G cellular networks and wireless metropolitan area networks(WMANs).

• Journal of KIISE:Computer Systems and Theory
• /
• v.30 no.9
• /
• pp.512-522
• /
• 2003

For a time-multiplexed FPGA, a circuit is partitioned into several subcircuits, so that they temporally share the same physical FPGA device by hardware reconfiguration. In these architectures, all the hardware reconfiguration information called contexts are generated and downloaded into the chip, and then the pre-scheduled context switches occur properly and timely. Typically, the size of the chip required to implement the circuit depends on both the maximum number of the LUT blocks required to implement the function of each subcircuit and the maximum number of micro-registers to store results over context switches in the same time. Therefore, many partitioning or synthesis methods try to minimize these two factors. In this paper, we present a new estimation technique to find the lower bound on the number of micro-registers which can be obtained by any synthesis methods, respectively, without performing any actual synthesis and/or design space exploration. The lower bound estimation is very important in sense that it greatly helps to evaluate the results of the previous work and even the future work. If the estimated lower bound exactly matches the actual number in the actual design result, we can say that the result is guaranteed to be optimal. In contrast, if they do not match, the following two cases are expected: we might estimate a better (more exact) lower bound or we find a new synthesis result better than those of the previous work. Our experimental results show that there are some differences between the numbers of micro-registers and our estimated lower bounds. One reason for these differences seems that our estimation tries to estimate the result with the minimum micro-registers among all the possible candidates, regardless of usage of other resources such as LUTs, while the previous work takes into account both LUTs and micro-registers. In addition, it implies that our method may have some limitation on exact estimation due to the complexity of the problem itself in sense that it is much more complicated than LUT estimation and thus needs more improvement, and/or there may exist some other synthesis results better than those of the previous work.

• Journal of Intelligence and Information Systems
• /
• v.24 no.2
• /
• pp.221-241
• /
• 2018

Deep learning is getting attention recently. The deep learning technique which had been applied in competitions of the International Conference on Image Recognition Technology(ILSVR) and AlphaGo is Convolution Neural Network(CNN). CNN is characterized in that the input image is divided into small sections to recognize the partial features and combine them to recognize as a whole. Deep learning technologies are expected to bring a lot of changes in our lives, but until now, its applications have been limited to image recognition and natural language processing. The use of deep learning techniques for business problems is still an early research stage. If their performance is proved, they can be applied to traditional business problems such as future marketing response prediction, fraud transaction detection, bankruptcy prediction, and so on. So, it is a very meaningful experiment to diagnose the possibility of solving business problems using deep learning technologies based on the case of online shopping companies which have big data, are relatively easy to identify customer behavior and has high utilization values. Especially, in online shopping companies, the competition environment is rapidly changing and becoming more intense. Therefore, analysis of customer behavior for maximizing profit is becoming more and more important for online shopping companies. In this study, we propose 'CNN model of Heterogeneous Information Integration' using CNN as a way to improve the predictive power of customer behavior in online shopping enterprises. In order to propose a model that optimizes the performance, which is a model that learns from the convolution neural network of the multi-layer perceptron structure by combining structured and unstructured information, this model uses 'heterogeneous information integration', 'unstructured information vector conversion', 'multi-layer perceptron design', and evaluate the performance of each architecture, and confirm the proposed model based on the results. In addition, the target variables for predicting customer behavior are defined as six binary classification problems: re-purchaser, churn, frequent shopper, frequent refund shopper, high amount shopper, high discount shopper. In order to verify the usefulness of the proposed model, we conducted experiments using actual data of domestic specific online shopping company. This experiment uses actual transactions, customers, and VOC data of specific online shopping company in Korea. Data extraction criteria are defined for 47,947 customers who registered at least one VOC in January 2011 (1 month). The customer profiles of these customers, as well as a total of 19 months of trading data from September 2010 to March 2012, and VOCs posted for a month are used. The experiment of this study is divided into two stages. In the first step, we evaluate three architectures that affect the performance of the proposed model and select optimal parameters. We evaluate the performance with the proposed model. Experimental results show that the proposed model, which combines both structured and unstructured information, is superior compared to NBC(Naïve Bayes classification), SVM(Support vector machine), and ANN(Artificial neural network). Therefore, it is significant that the use of unstructured information contributes to predict customer behavior, and that CNN can be applied to solve business problems as well as image recognition and natural language processing problems. It can be confirmed through experiments that CNN is more effective in understanding and interpreting the meaning of context in text VOC data. And it is significant that the empirical research based on the actual data of the e-commerce company can extract very meaningful information from the VOC data written in the text format directly by the customer in the prediction of the customer behavior. Finally, through various experiments, it is possible to say that the proposed model provides useful information for the future research related to the parameter selection and its performance.