• Phonetics and Speech Sciences
• /
• 제12권2호
• /
• pp.29-37
• /
• 2020

Recently, the neural network-based deep learning algorithm has dramatically improved performance compared to the classical Gaussian mixture model based hidden Markov model (GMM-HMM) automatic speech recognition (ASR) system. In addition, researches on end-to-end (E2E) speech recognition systems integrating language modeling and decoding processes have been actively conducted to better utilize the advantages of deep learning techniques. In general, E2E ASR systems consist of multiple layers of encoder-decoder structure with attention. Therefore, E2E ASR systems require data with a large amount of speech-text paired data in order to achieve good performance. Obtaining speech-text paired data requires a lot of human labor and time, and is a high barrier to building E2E ASR system. Therefore, there are previous studies that improve the performance of E2E ASR system using relatively small amount of speech-text paired data, but most studies have been conducted by using only speech-only data or text-only data. In this study, we proposed a semi-supervised training method that enables E2E ASR system to perform well in corpus in different domains by using both speech or text only data. The proposed method works effectively by adapting to different domains, showing good performance in the target domain and not degrading much in the source domain.

• KIPS Transactions on Software and Data Engineering
• /
• 제5권7호
• /
• pp.319-326
• /
• 2016

Given the attributes of mobile apps that shorten the time to make purchase decisions while enabling easy purchase cancellations, usability can be regarded to be a highly prioritized quality attribute among the diverse quality attributes that must be provided by mobile apps. With that backdrop, mobile app developers have been making great effort to minimize usability hampering elements that degrade the merchantability of apps in many ways. Most elements that hamper the convenience in use of mobile apps stem from those potential errors that occur when GUIs are designed. In our previous study, we have proposed a technique to analyze the usability of mobile apps using user behavior logs. We proposes a technique to detect usability hampering elements lying dormant in mobile apps' GUI models by expressing user behavior logs with finite state models, combining user behavior models extracted from multiple users, and comparing the combined user behavior model with the expected behavior model on which the designer's intention is reflected. In addition, to reduce the burden of the repeated test operations that have been conducted by existing developers to detect usability errors, the present paper also proposes a mobile usability error detection automation tool that enables automatic application of the proposed technique. The utility of the proposed technique and tool is being discussed through comparison between the GUI issue reports presented by actual open source app developers and the symptoms detected by the proposed technique.

• KIPS Transactions on Software and Data Engineering
• /
• 제2권2호
• /
• pp.113-118
• /
• 2013

The current GPU virtualization techniques incur large overheads when executing application programs mainly due to the fine-grain time-sharing scheduling of the GPU among multiple Virtual Machines (VMs). Besides, the current techniques lack of portability, because they include the APIs for the GPU computations in the VM monitor. In this paper, we propose a low overhead and high performance GPU virtualization approach on a heterogeneous HPC system based on the open-source Xen. Our proposed techniques are tailored to the bio applications. In our virtualization framework, we allow a VM to solely occupy a GPU once the VM is assigned a GPU instead of relying on the time-sharing the GPU. This improves the performance of the applications and the utilization of the GPUs. Our techniques also allow a direct pass-through to the GPU by using the IOMMU virtualization features embedded in the hardware for the high portability. Experimental studies using microbiology genome analysis applications show that our proposed techniques based on the direct pass-through significantly reduce the overheads compared with the previous Domain0 based approaches. Furthermore, our approach closely matches the performance for the applications to the bare machine or rather improves the performance.

• Archives of Plastic Surgery
• /
• 제36권5호
• /
• pp.629-636
• /
• 2009

Purposes: Wide scars occurring on the lower face and neck are a source of both functional and esthetic problems. Consequently, we can use skin grafts, pedicled flaps, free flaps, and tissue expansion for the reconstruction of this area. Compared with other reconstruction techniques, tissue expansion is advantageous in that it enables the maintenance of a color and texture similar to that of the adjacent tissue. However, the conventional method of tissue expansion has been reported to lead to an unnatural cervicomental angle and to the deformity of adjacent structures. We have therefore made efforts to prevent these problems through the use of several operative procedures. Methods: Forty-one patients with lower facial and cervical scars underwent tissue expansion. The tissue expansion was performed using a rectangular-shaped Nagosil$^{(R)}$ tissue expansion device. On insertion of the tissue expander, the intermediate area of superficial fat layer was dissected and then the tissue expander was inserted to make a flap that was as thin as possible. In advancement of the flap, a capsule-formed by the tissue expander-was used for the interrupted fixed suture of the flap to the fascia of the platysma muscle of the neck. This procedure was performed multiple times and also performed between the flap and the periosteum of the mandible, such that the tension was removed during the suture of the flap margin. Finally, the patients were fitted with a Jobst$^{(R)}$ facial garment in order to stabilize the operation site at least twelve months. Results: The most prevalent location of the scar was the cheek (15 cases), followed by the chin in 14 cases and the neck in 12 cases. The mean size of scar was $55.7{\pm}39.4cm^2$. Conclusions: Using our procedures, we have experienced no significant deformities and have also achieved a more natural cervicomental angle in the patients.

• Current Research on Agriculture and Life Sciences
• /
• 제32권3호
• /
• pp.111-117
• /
• 2014

Phytophthora blight caused by Phytophthora capsici Leonian is a dangerous disease threatening pepper growers worldwide. The efficacy of chemical control is generally low as the pathogen is soil-borne and rapidly spread by zoospores during the rainy season. Thus, based on the demand for resistant varieties, various good resistant sources, such as CM334, AC2258, and PI201234, have been reported and their inheritance of resistance studied by many different authorities. However, the mode of inheritance remains unclear, as 1 or 2 independent dominant genes, 3 genes, or multiple genes have all been reported as responsible for resistance. Recently, QTL mappings of the gene factors for resistance have been reported, and molecular markers for resistance used in breeding programs. With the release of many resistant commercial hybrid cultivars, differentiation of pathotypes of the pathogen is attracting interest among breeders and plant pathologists. Various authorities have already classified the pathogen strains into different races according to the inter-action between resistant host plants, including the source of resistance, such as CM334 and PI201234, and resistant commercial varieties and P. capsici isolates. However, no standard differential host sets have yet been established, so the results are good only for the pathogen strains used in the experiments. Thus, for breeding varieties with durable resist-ance, it is important to introduce resistance from different sources and use diverse local pathogen strains collected in the target area for distribution in a breeding program.

• Korean Journal of Remote Sensing
• /
• 제22권2호
• /
• pp.141-151
• /
• 2006

UV-visible absorption measurement techniques using several horizone viewing directions in addition to the traditional zenith-sky pointing have been recently developed in ground-based remote sensing of atmospheric constituents. The spatial distribution of various trace gases close to the instrument can be derived by combing several viewing directions. Multi-axis differential optical absorption spectroscopy (MAX-DOAS) technique, one of the remote sensing techniques for air quality measurements, uses the scattered sunlight as a light source and measures it at various elevation angles (corresponding to the viewing directions) by sequential scanning with a stepper motor. A MAX-DOAS system developed by GIST/ADEMRC has been applied to measuring trace gases in urban air and plumes of the volcano and fossil fuel power plant in January, May, and October 2004, respectively. MAX-DOAS spectra were analyzed to identify and quantify $SO_2,\;NO_2,\;BrO,\;and\;O_4$ (based on Slant Column Densities, SCD) in the urban air, volcanic plume, and fossil fuel power plant utilizing theirs specific structured absorption features in the UV-visible region. Vertical scan through the multiple elevation angles was performed at different directions perpendicular to the plume dispersion to retrieve cross-sectional distribution of $SO_2\;or\;NO_2$ in the plumes of the volcano and fossil fuel power plant. Based on the estimated cross sections of the plumes the mixing ratios were estimated to 580 $SO_2$ ppbv in the volcanic Plume, and 337 $NO_2\;and\;227\;SO_2$ ppbv in the plume of the fossil fuel power plant, respectively.

• Korean Journal of Plant Taxonomy
• /
• 제46권4호
• /
• pp.371-377
• /
• 2016

The succulent herbaceous perennial Sedum sarmentosum commonly grows in the southern part of the Korean Peninsula. It is a species native to China, most likely introduced into Korea due to its edible and medicinal uses. If plants were introduced from a single source, we would expect no or low levels of genetic variation in Korean populations. Alternatively, if plants were introduced from multiple sources, we would expect, in contrast, high levels of genetic diversity. To test which is more likely, we surveyed the degree of allozyme variation in ten populations of this species from southern Korea. We found that S. sarmentosum was monomorphic at all fifteen allozyme loci. However, two congeners (S. polytrichoides and S. kamtschaticum) and two related species (Hylotelephium ussuriense and H. verticillatum) maintain moderate to high levels of genetic diversity ($H_e=0.144$, 0.203, 0.201, and 0.204, respectively). We suggest that southern Korean populations of S. sarmentosum likely descended from a single introduction of a few plants and then became naturalized exclusively via vegetative spreading (as plants in Korea, but also as occurs in other parts of its native and naturalized range, are sterile).

• Journal of the Institute of Convergence Signal Processing
• /
• 제19권4호
• /
• pp.153-161
• /
• 2018

This paper describes the performance improvement of Software (SW) digital filter using GPU (Graphical Processing Unit). The previous developed SW digital filter has a problem that it operates on a CPU (Central Processing Unit) basis and has a slow speed. The GPU was introduced to filter the data of the EAVN (East Asian VLBI Network) observation to improve the operation speed and to process data with other stations through filtering, respectively. In order to enhance the computational speed of the SW digital filter, NVIDIA Titan V GPU board with built-in Tensor Core is used. The processing speed of about 0.78 (1Gbps, 16MHz BW, 16-IF) and 1.1 (2Gbps, 32MHz BW, 16-IF) times for the observing time was achieved by filtering the 95 second observation data of 2 Gbps (512 MHz BW, 1-IF), respectively. In addition, 2Gbps data is digitally filtered for the 1 and 2Gbps simultaneously observed with KVN (Korean VLBI Network), and compared with the 1Gbps, we obtained similar values such as cross power spectrum, phase, and SNR (Signal to Noise Ratio). As a result, the effectiveness of developed SW digital filter using GPU in this research was confirmed for utilizing the data processing and analysis. In the future, it is expected that the observation data will be able to be filtered in real time when the distributed processing optimization of source code for using multiple GPU boards.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• 제18권6호
• /
• pp.133-137
• /
• 2018

Arduino is a small microcomputer that is used in a variety of industry fields and especially is widely used as an open source hardware IoT device. The multi-tasking method of Arduino is divided into super loop timing and RTOS thread method. The super loop timing method is simple and easy to understand. However, when one task is long, it affects the execution of the next task. In addition, RTOS threading has the advantage of being able to run without being influenced by other work time. However, Arduino, a small microcomputer, has a disadvantage in that, when the number of threads increases, the context switching time of the thread causes additional time not included in the super loop timing method have. In this paper, we use Arduino Uno R3 and FreeRTOS to analyze these different features, and the task for the experiment is to send 8000 digital signals to the built-in LED port. If two tasks of the same size are executed, the super loop method executes 3 ms faster than FreeRTOS multitasking. If multiple tasks are executed simultaneously, superloop type task is sequential execution and difference in execution time between first task and last task is large. FreeRTOS method can be executed concurrently, but execution time delay of about 30 ms occurs in context switching time.

• Korean Journal of Environmental Biology
• /
• 제36권4호
• /
• pp.507-516
• /
• 2018

The closed chamber method, which is one of the most commonly used method for measuring greenhouse gases produced in rice paddy fields, has limitations in measuring dynamic $CH_4$ flux with spatio-temporal constrains. In order to deal with the limitation of the closed chamber method, some studies based on open-path of eddy covariance method have been actively conducted recently. The aim of this study was to compare the $CH_4$ fluxes measured by open-path and closed chamber method in the paddy rice fields. The open-path, one of the gas ($CO_2$, $CH_4$ etc.) analysis methods, is technology where a laser beam is emitted from the source passes through the open cell, reflecting multiple times from the two mirrors, and then detecting. The $CH_4$ emission patterns by these two methods during rice cultivation season were similar, but the total $CH_4$ emission measured by open-path method were 31% less than of the amount measured by closed chamber. The reason for the difference in $CH_4$ emission was due to overestimation by closed chamber and underestimation by open-path. The closed chamber method can overestimate $CH_4$ emissions due to environmental changes caused by high temperature and light interruption by acrylic partition in chamber. On the other hand, the open-path method for eddy covariance can underestimate its emission because it assumes density fluctuations and horizontal homogeneous terrain negligible However, comparing $CH_4$ fluxes at the same sampling time (AM 10:30-11:00, 30-min fluxes) showed good agreements ($r^2=0.9064$). The open-path measurement technique is expected to be a good way to compensate for the disadvantage of the closed chamber method because it can monitor dynamic $CH_4$ fluctuation even if data loss is taken into account.