• 한국지하수토양환경학회:학술대회논문집
• /
• 한국지하수토양환경학회 2003년도 총회 및 춘계학술발표회
• /
• pp.25-39
• /
• 2003

Risk and exposure assessment for subsurface environment is very important for both aspects of health and environmental protection as well as making decision of remedial goal for engineering activities. Exposure due to hazardous chemicals in the subsurface environment is essential to assess risk lev121 to individual person, especially from soil and groundwater environmental media. In this paper, the status of soil and groundwater contamination is presented to discuss on the problem for environmental risk assessment. The methodologies of fate and exposure models are also discussed by conducting the case studies of exposure assessment for heavy metals, organic compounds, and dioxin compounds. In addition, the structure of exposure models and available data for model calculation are examined to make clear more realistic exposure scenarios and the application to the practical environmental issues. Three kinds of advanced remediation techniques for soil and groundwater contamination are described in this paper, The most practical method for VOCs is the bio-remediation technique in which biological process due to consortium of microorganisms can be applied. For more effective remediation of soil contaminated by heavy metals we have adopted the soil flushing technique and clean-up system using electro-kinetic method. We have also developed the advanced techniques of geo-melting method for soil contaminated by DXNs and PCB compounds. These techniques are planed to introduce and to apply for a lot of contaminated sites in Japan.

• 환경영향평가
• /
• 제8권1호
• /
• pp.61-70
• /
• 1999

The purpose of this study is to understand an appreciation and a perceptive structure of "keijeok" image by Semantic Differential Technique and Factor Analysis. The data used in this study was obtained by the questionnaire survey carried out in Pusan metropolitan city. 15 adjective pairs in the survey were evaluated by the Semantic Differential scales graded 7 ranges from 1(very good) to 7(very bad). A total of 452 samples were collected by the survey and analyzed for this study. The results are as follows; First, 15 variables comprehended to "keijeok" image were estimated as a positive conception(LT 4.0). What's more, residents may perceive "keijeok" image as intangible and aesthetic aspect such as "fresh", "pleasant", "clean". Second, the result of factor analysis shows that factor I which express the major conceptual meaning of "keijeok" image tends to have intangible or aesthetic adjective pairs rather than concrete, whereas factor II which has the weaker meaning compared with factor I may represent a functional aspect of "keijeok" image. It can explain that the perceptive structure of "keijeok" image may be largely influenced by subjective sense, then added or concreted with objective conception or environmental situation. The results can be considered as an important matter which should be reflected at the stage of environmental planning for people's amiable and desirable place.

• 한국통신학회논문지
• /
• 제37권6C호
• /
• pp.492-496
• /
• 2012

본 논문은 잡음이 많이 존재할 경우 특징 보상 기법들의 불완전한 추정 방법으로 인하여 발생할 수 있는 불확실성 정보를 음성 인식의 디코딩에 반영해 줌으로써 좀 더 인식 성능을 향상시킬 수 있는 방법에 대한 연구이다. 기존의 특징 보상 기법들은 현재 시간에서의 깨끗한 특징 파라미터를 추정하는 단일점 추정 기법들이 대부분이다. 하지만 낮은 SNR 환경에서의 잘못된 추정 파라미터들이 음성 인식 엔진의 입력으로 사용될 경우 성능이 저하되기 때문에 추정된 파라미터의 불확실성 정보를 이용하여 디코딩을 해주면 추정 오류를 보완해줄 수 있다. 본 논문에서는 대표적인 Aurora-2 DB를 활용하여 적용된 기법의 성능 향상을 확인한다.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제13권8호
• /
• pp.4285-4299
• /
• 2019

Adversarial attack is a technique that causes a malfunction of classification models by adding noise that cannot be distinguished by humans, which poses a threat to a deep learning model. In this paper, we propose an efficient method to detect adversarial images using Gaussian process regression. Existing deep learning-based adversarial detection methods require numerous adversarial images for their training. The proposed method overcomes this problem by performing classification based on the statistical features of adversarial images and clean images that are extracted by Gaussian process regression with a small number of images. This technique can determine whether the input image is an adversarial image by applying Gaussian process regression based on the intermediate output value of the classification model. Experimental results show that the proposed method achieves higher detection performance than the other deep learning-based adversarial detection methods for powerful attacks. In particular, the Gaussian process regression-based detector shows better detection performance than the baseline models for most attacks in the case with fewer adversarial examples.

• 한국지구물리탐사학회:학술대회논문집
• /
• 한국지구물리탐사학회 2006년도 공동학술대회 논문집
• /
• pp.253-258
• /
• 2006

한반도는 고도의 산업화와 도시화로 인해 깨끗한 MT 및 GDS 자료를 획득하기 매우 어렵다. 이러한 환경에서는 적당한 신호 대 잡음비를 가지는 자료를 선별하는 전처리 과정이 필수적이다. 본 연구에서는 전자기장의 분광에너지에 기반한 자동 전처리 과정(RMP)을 한반도의 현실적인 상황을 고려하여 수정하였고, 이를 한반도 내륙에서 획득된 총 7 측점의 MT 자료에 적용하였다. 전반적으로, RMP 기법은 주기 10초 대역 부근에서 주파수에 따른 전기비저항과 위상의 연결성을 상당히 향상시켰다. 또한 RMP 기법에 의한 결과는 수작업으로 편집된 기존의 처리 결과와도 거의 유사였고, 특히 전기 비저항의 연결성면에서 그 우수성이 확인되었다.

• 한국안전학회지
• /
• 제23권3호
• /
• pp.1-7
• /
• 2008

As a clean fuel, LNG or LPG is widely used in large restaurants or plants instead of coal. These fuels are easy to use and cheap. However, they are inflammable very much and easily exploded at the worst. So there are many laws and regulations legislated for the handling and usage of the gas facilities. Nevertheless, very harsh explosions occurred in several domestic or foreign places. In this paper a gas detection and alarm system was developed for large facilities. A new concept for connecting many gas detectors was proposed and based upon that a system including hardware and software have been developed and tested. Although up to 16 gas detectors apart 1 km from main controller can be connected currently, this system can be easily expanded for more detectors because each gas detector receiver communicates with main controller using industrial standard RS-485 multi drop technique. Furthermore several additive functions was included for the installation and maintenance convenience.

• 말소리와 음성과학
• /
• 제6권1호
• /
• pp.77-83
• /
• 2014

This paper proposes a speaker and environment clustering method in order to overcome the degradation of the speech recognition performance caused by various noise and speaker characteristics. In this paper, instead of using the distance between Gaussian mixture model (GMM) weight vectors as in the Google's approach, the distance between the adapted mean vectors based on the modified maximum a posteriori (MAP) adaptation is used as a distance measure for vector quantization (VQ) clustering. According to our experiments on the simulation data generated by adding noise to clean speech, the proposed clustering method yields error rate reduction of 10.6% compared with baseline speaker-independent (SI) model, which is slightly better performance than the Google's approach.

• 한국신재생에너지학회:학술대회논문집
• /
• 한국신재생에너지학회 2005년도 춘계학술대회
• /
• pp.384-387
• /
• 2005

Fuel cell has been well known as a clean alternative power for vehicles. Recently, an experimental technique has been developed measurement of species and distributions by using gas chromatograph. In this study, cathode channel oxygen distributions as various conditions were investigated using gas chromatograph and cell visualization. And discussed relation between flooding and oxygen concentrations. As a result of experiment, oxygen consumpt ion is affected wi th flooding. Flooding is observed in channel near hydrogen inlet, and oxygen consumption is low at that region.

• 한국정밀공학회지
• /
• 제17권4호
• /
• pp.209-219
• /
• 2000

A method is developed to evaluate machining process and to determine the optimal machining conditions considering the environmental effects. The method Is based on the evaluation attributes from the general LCA programs and the analysis technique of AHP from HHS. To assist the analysis. the mass models of cutting energy, tools, and fluids are developed. The models may be used for both quantitative prediction of the uses and disposed masses of materials and optimization of the machining conditions. The algorithm with the mass models is applied to the milling process planning. The process to survey the environmental data, calculate the used mass, and evaluate the alternatives is demonstrated. This demonstration illustrates the of the change of process conditions of the decision making.

• 한국환경보건학회지
• /
• 제27권3호
• /
• pp.107-112
• /
• 2001

This work reports the application of a hollow fiber module(HFM) for Cr(VI) extraction from heavy metal salts mixed solution by using microporous hydrophobic hollow fiber module. In HFM configuration, the organic extraction used for the extraction of Cr(VI) was di-(2-ethyl hexyl) phosphoric acid(D2EHPA) diluted with n-heptane. The study of HFM includes the influence of hydrodynamic and chemical condition, i.e., the flow rate of feed solution, the time of reactive extraction, the concentration of feed solution, and the pH of aqueous phase solutions. Several experiments with synthetic solution of different mixed components system of Cr(VI) solutions established optimum condition to achieve a clean separation of Cr(VI). It was possible to separate Cr(VI) in the presence of metal salts mixed solution, such as Zn(II), Ni(II), Cu(II), and Cd(II) using the HFM technique.