• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.12 no.10
• /
• pp.4930-4951
• /
• 2018

The deep learning based speech enhancement has shown considerable success. However, it still suffers performance degradation under mismatch conditions. In this paper, an adaptation method is proposed to improve the performance under noise mismatch conditions. Firstly, we advise a noise aware training by supplying identity vectors (i-vectors) as parallel input features to adapt deep neural network (DNN) acoustic models with the target noise. Secondly, given a small amount of adaptation data, the noise-dependent DNN is obtained by using $L_2$ regularization from a noise-independent DNN, and forcing the estimated masks to be close to the unadapted condition. Finally, experiments were carried out on different noise and SNR conditions, and the proposed method has achieved significantly 0.1%-9.6% benefits of STOI, and provided consistent improvement in PESQ and segSNR against the baseline systems.

• Proceedings of the KIEE Conference
• /
• 1995.11a
• /
• pp.268-270
• /
• 1995

In crane drives, DC motor has been most widely used due to simple control characteristic and favorable transient behavior. Nowadays, however, the squirrel cage induction motor is known as an attractive candidate due to elimination of all sliding electrical contacts, resulting in an exceedingly simple and rugged construction. Especially, in hoist application, the smooth torque control and four quadrant operation are essential. In this paper, an operation of dual inverters with common DC bus fed by vector controlled induction motor is described. Single DSP is employed as a main processor to control dual inverters and communicates each input/output signal with PLC. As well as giving a detailed expression, full simulation and experimental results are presented.

• Ocean and Polar Research
• /
• v.31 no.4
• /
• pp.305-318
• /
• 2009

A marine ranching project in Tongyoung was established in 1998, lasting 9 years to 2006. Project activities included the deployment of artificial reefs, the release of young fishes like jacopever and rockfish, and input/output control for specific marine ranching areas in Tongyoung. This report focuses on the economic feasibility of the project in hindsight. Analysis concentrates on three aspects; (a) direct economic benefits, such as increasing effects of fisheries income and savings in harvesting costs, (b) indirect benefits, including increasing effects of recreational fishing and saving R&D costs, and (c) costs, including releasing and purchasing costs of artificial reef and juvenile fish, R&D costs, maintenance costs and harvesting costs. Results show that NPV=4.7 billion won, IRR=8.55% and B/C ratio=1.286 under Scenario 1, which considers the saving effects of R&D costs, and NPV=0.9 billion won, IRR=6.03% and B/C ratio=1.11 under Scenario 2, which does not consider the saving effects of R&D costs, based on 5.5% of the social rate of discount. According to sensitivity analysis, the economic feasibility is very sensitive to the recapture rate.

• Proceedings of the Korean Institute of Electrical and Electronic Material Engineers Conference
• /
• 2004.05b
• /
• pp.19-22
• /
• 2004

The resonant planar transformer, which had power capacity of 300 W, input voltage of 220 V, output voltage of 15 V, and switching frequency of 500 kHz, was designed and manufactured by using the planar core with large effective area and the flat copper lead frames for miniaturization and high efficiency of the switching mode power supply (SMPS). As well as, a resonant converter equipped with the above mentioned planar transformer was manufactured and electromagnetic characteristics were investigated. The numerical value of turns for 1st and 2nd winding were 12 and 2 respectively. The self inductance of 1st winding was 33.2 ${\mu}H$, very low leakage inductance of 1.27 ${\mu}H$, and the coupling factor of 0.98 were obtained at switching frequency of 300 kHz. The high efficiency of 88.21 % for the SMPS equipped with planar transformer was obtained at power capacity of 300 W.

• Proceedings of the KIEE Conference
• /
• 1993.07a
• /
• pp.227-230
• /
• 1993

Recently, the development of novel control methodology enables us to improve the performance of AC-machine drives by using pulse width modulation (PWM) technique. Usually, the dynamic characteristic of induction motor (IM) has been represented by the 5-th order nonlinear differential equation. This dynamics, however, can be reduced to 3-rd order dynamics by applying direct control of IM input current. This methodology concludes that it is much easier to control IM by means of the field-oriented methods employing the current controller. Therefore a precise current control is crucial to achieve a high control performance both in dynamic and steady state operations. This paper presents an adaptive fuzzy current controller with artificial neural network (ANN) for field-oriented controlled IM. This new control structure is able to adaptively minimize a current ripple while maintaining constant switching frequency. Especially the proposed controller employs neuro-computing philosophy as well as adaptive learning pattern recognizing principles with respect to variations of the system parameters. The proposed approach is applied to the IM drive system, and its performance is tested through various simulations. Simulation results show that the proposed system, compared among several known classical methods, has a superb performance.

• Journal of the Society of Cosmetic Scientists of Korea
• /
• v.16 no.1
• /
• pp.47-63
• /
• 1990

Image analyser was used to understand the condition of skin surface and to evaluate the efficacy of cosmetic treatment. It was unsatisfactory to analyse skin surface structure although several methods using image analyser had been presented. We developed the new system composed of image analyser and Tandem Scanning Reflected Light Microscope (TSRLM) having the remarkable optical sectioning property as image input device. By using this new system, we quantitatively measured the change of skin surface, the depth and width of furrow in micron unit, resulted by cosmetic treatments. And also three dimensional image of skin was reconstructed with serial sectioned images, which were captured through TSRLM, for better understanding of the effect of cosmetic treatment. It was found that skin relief was more easily understood and the change of skin surface caused by cosmetic treatment was more accurately measured by using this system. In addition, we was also aware of the possibility of in vivo direct measurement of skin furrow without replica. It was conceivable that our system could be applicable for study of cosmetic effects further.

• Journal of Digital Convergence
• /
• v.19 no.5
• /
• pp.31-39
• /
• 2021

This study collected 1,870 English news articles related to research and development of autonomous vehicles in order to identify various issues emerging in the research and development process of autonomous vehicles at home and abroad, and conducted topic modeling after data pre-processing. As a result of topic modeling, we extracted 20 topics, and we performed naming operations for topics and interpreted their meanings. A logical model for autonomous vehicle research and development projects was presented in response to the R&D process of input, activity, output, and outcome of derived topics. The analysis results of this study will be used as basic data to accurately determine the progress of domestic and foreign self-driving car research and development projects and prepare for the rapidly changing technology development.

• Journal of Korea Multimedia Society
• /
• v.22 no.9
• /
• pp.1000-1010
• /
• 2019

Image semantic segmentation and dehazing are key tasks in the computer vision. In recent years, researches in both tasks have achieved substantial improvements in performance with the development of Convolutional Neural Network (CNN). However, most of the previous works for semantic segmentation assume the images are captured in clear weather and show degraded performance under hazy images with low contrast and faded color. Meanwhile, dehazing aims to recover clear image given observed hazy image, which is an ill-posed problem and can be alleviated with additional information about the image. In this work, we propose a deep multi-task network for simultaneous semantic segmentation and dehazing. The proposed network takes single haze image as input and predicts dense semantic segmentation map and clear image. The visual information getting refined during the dehazing process can help the recognition task of semantic segmentation. On the other hand, semantic features obtained during the semantic segmentation process can provide cues for color priors for objects, which can help dehazing process. Experimental results demonstrate the effectiveness of the proposed multi-task approach, showing improved performance compared to the separate networks.

• Proceedings of the KIEE Conference
• /
• 1996.11a
• /
• pp.123-126
• /
• 1996

In this paper, optimal modulation controller is designed to improve the stability of A.C. and A.C.-D.C. power system, and optimal theory is applied to select optimal modulation controller input signal Optimal modulation controller for speed governor and exciter controller system is constructed in A.C. power system, while the controller is constructed to the both control systems like AC. power system, considering ACR-AVR, APR-$A{\gamma}R$ as the control method of direct current system. It is considered that the stability of A.C. power system only and A.C.-D.C. power system against load fluctuations and disturbances under case of optimal modulation control.

• Journal of Technology Innovation
• /
• v.22 no.4
• /
• pp.165-198
• /
• 2014

The national R&D performance evaluation system in Korea has been influenced by 'New Public Management' since the 1980s. The 1997 IMF crisis in Korea has necessitated the overhaul of S&T policy strategies in Korea. Reflecting this, effective distribution and utilization of S&T investment and resources has emerged as a crucial policy agenda, while expansionary investment in S&T maintained. During the period of 1999-2005, the R&D evaluation has been more focused on the input side and execution processes than on the results of R&D. Evaluation results were to be reflected in the budgeting of national R&D, but there were some disputes over how to make use of evaluation results between the budgeting ministry and the evaluation agency partly because they were two separate entities. After the advent of the new government(1998-2003) which advocated 'science and technology-oriented society', the national R&D evaluation system has evolved through a new legislation, the Performance Assessment Act, and the establishment of Science & Technology Innovation Agency to build up the unique framework for national R&D evaluation differentiated from the assessment system for general government investment programs. Most recently, due to 'Comprehensive Improvement Plan' in 2013, various components of national R&D evaluation seem to move on a new evolutionary track. For example, different types of evaluation are being developed and tried in accordance with differing evaluation goals, and the individual ministries and the research councils has got more enhanced autonomy and responsibility regarding R&D evaluation. So-called 'tailored' evaluation methods, taking into consideration the characteristics of programs and institutes / organizations to be evaluated, are now being tried. Competences of evaluation experts and agencies has also to be improved and strengthened.