• 대한음성학회지:말소리
• /
• 제56호
• /
• pp.207-223
• /
• 2005

The goal of language model adaptation is to improve the background language model with a relatively small adaptation corpus. This study presents a language model adaptation technique where additional text data for the adaptation do not exist. We propose the information retrieval (IR) technique with N-gram language modeling to collect the adaptation corpus from baseline text data. We also propose to use a dynamic language model interpolation coefficient to combine the background language model and the adapted language model. The interpolation coefficient is estimated from the word hypotheses obtained by segmenting the input speech data reserved for held-out validation data. This allows the final adapted model to improve the performance of the background model consistently The proposed approach reduces the word error rate by $13.6\%$ relative to baseline 4-gram for two-hour broadcast news speech recognition.

• 산업공학
• /
• 제14권1호
• /
• pp.79-83
• /
• 2001

In this paper, we propose a simple method to reduce categorical variables into smaller, but significant numbers, and also demonstrate how the proposed method can be applied to the problem of reduction that empirical research often faces in the course of data processing. For the purpose, we introduce a concept of pseudo-correlation coefficient to make it possible to use factor analysis (FA) as a tool for reducing variables. The main idea of the concept is to deal with the measures of association of categorical variables in the sense of the concept of Pearson's correlation coefficient in order to meet the input requirement of FA. Upon examination of existing measures that could play as pseudo-correlation coefficients, Cramer's V coefficient is selected for the best result among them. To show the detailed procedure of the proposed method, a specific demonstration with the data from 329 R&D projects conducted in 18 private laboratories in electric and electronics industry is presented.

• Food Science and Biotechnology
• /
• 제14권3호
• /
• pp.328-333
• /
• 2005

Finite difference model and dynamic thermal property evaluation system were developed to estimate convection heat transfer coefficient by modeling temperature-time profile of beef cube in continuous flow sterilizing system. As input parameters of the model, specific heat and thermal conductivity values of beef frankfurter meat were independently measured from 20 to $80^{\circ}C$. Convection heat transfer coefficient was estimated by comparing simulated and measured temperature-time profiles. Actual temperature-time profiles of meat cube were measured at flow rates of 15, 30, and 45 L/min and viscosities from 0 to 15 cp, and mean values of convection heat transfer coefficients ranged from 792 to $2107\;W/m^2{\cdot}K$. Convection heat transfer coefficient increased with increase in flow rate and decreased as viscosity increased.

• Geomechanics and Engineering
• /
• 제30권3호
• /
• pp.259-272
• /
• 2022

Rockburst is a dynamic, multivariate, and non-linear phenomenon that occurs in underground mining and civil engineering structures. Predicting rockburst is challenging since conventional models are not standardized. Hence, machine learning techniques would improve the prediction accuracies. This study describes decision based uncertainty models to predict rockburst in underground engineering structures using gradient boosting algorithms (GBM). The model input variables were uniaxial compressive strength (UCS), uniaxial tensile strength (UTS), maximum tangential stress (MTS), excavation depth (D), stress ratio (SR), and brittleness coefficient (BC). Several models were trained using different combinations of the input variables and a 3-fold cross-validation resampling procedure. The hyperparameters comprising learning rate, number of boosting iterations, tree depth, and number of minimum observations were tuned to attain the optimum models. The performance of the models was tested using classification accuracy, Cohen's kappa coefficient (k), sensitivity and specificity. The best-performing model showed a classification accuracy, k, sensitivity and specificity values of 98%, 93%, 1.00 and 0.957 respectively by optimizing model ROC metrics. The most and least influential input variables were MTS and BC, respectively. The partial dependence plots revealed the relationship between the changes in the input variables and model predictions. The findings reveal that GBM can be used to anticipate rockburst and guide decisions about support requirements before mining development.

• 대한의용생체공학회:의공학회지
• /
• 제20권4호
• /
• pp.435-442
• /
• 1999

본 연구에서는 PVC를 분류하기 위하여 웨이브렛 계수를 기반으로 하는 fuzzy-ART 네트워크를 설계하였다. 설계된 네트워크는 feature를 추출하는 부분과 fuzzy-ART 네트워크를 학습시키는 부분으로 구성된다. 우선 feature의 문턱치 구간을 설정하기 위하여 심전도 신호의 QRS를 검출하였고, 검출된 QRS는 Haar 웨이브렛을 이용한 웨이브렛 변환에 의해 주파수 분할하였다. 분할된 주파수 중에서 입력 feature를 추출하기 위하여 저주파 영역의 6번째 계수(D6)만을 선택하였다. D6신호는 입력 feature를 구성하기 위한 문턱치를 적용하여 fuzzy-ART 네트워크의 2진수 입력 feature로 전환하였고, PVC를 분류하기 위하여 fuzzy-ART네트워크를 학습시켰다. 본 연구의 성능을 평가하기 위하여 PVC가 포함된 MIT/BIH 데이터 베이스가 사용되었으며, fuzzy-ART 네트워크의 분류성능은 96.25%이었다.

• 터널과지하공간
• /
• 제2권2호
• /
• pp.212-223
• /
• 1992

The distinct element method(DEM) si well suited to the kinematic analysis of blocky rock masses. Two distinctive problems, a rock avalache and tunnel in jointed rock masses, are chosen to apply the DEM which is based on perfectly rigid behaviour of blocks. Investigated for both problems are the effects of the input parameters such as contact stiffnesses, friction coefficient and damping property. Using various types of models of the avalanche and tunne, an extensive parametric study is done to gain experiences in the method, and then to alleviate difficulties in determining parameter values suitable for a given problem. The coefficient of frictio has significant effects on all aspects of avalanche motion(travel distance, velocity and travel time), while the stiffnesses affect the rebounding and jumping motions after collision. The motion predicted by the models having single and mutiple blocks agrees well to the observations reported on the actual avalache. For the tunnel problem, the behaviour of the key block in an example tunnel is compared by testing values of the input parameters. The stability of the tunnel is dependent primarily on the friction coefficient, while the stiffness and damping properties influence the block velocity. The kinematic stability of a tunnel for underground unclear waste repository is analyzed using the joint geometry data(orientation, spacing and persistence) occurred in a tailrace tunnel. Allowing a small deviation to the mean orientation results in different modes of failure of the rock blocks around the tunnel. Of all parameters tested, the most important to the stability of the tunnel in blocky rock masses are the geometry of the blocks generated by mapping the joint and tunnel surfaces in 3-dimensions and also the friction coefficient of the joints particularly for the stability of the side walls.

• 한국콘텐츠학회논문지
• /
• 제21권12호
• /
• pp.363-374
• /
• 2021

본 연구는 Asian development bank/ ERCD가 2021년 발표한 2019년 산업연관표를 활용하여 광산업이 몽골 국민 경제에 얼마만큼 기여하는지를 분석하여 몽골 경제의 특성을 파악하고 향후 몽골 광산업의 발전을 위한 정책 수립과 몽골 경제 활성화 방안에 참고자료로 활용될 수 있게 하는 데에 목적이 있다. 본 연구를 위해 몽골 경제를 35개 산업으로 분류하여 국가 경제 기여도를 분석하였다. 분석 결과 몽골 광산업의 총생산 유발액은 38,418백만 달러, 생산유발계수는 열 합계는 1.473, 감응도계수는 1.696, 부가가치유발계수는 0.707, 생산유발계수는 1.473 로 나타났다. 몽골 광산업은 다른 산업보다 생산유발효과가 높고, 다른 산업을 이끌어가는 전략산업으로써 발전 잠재력이 큼을 알 수 있다.

• 해양환경안전학회:학술대회논문집
• /
• 해양환경안전학회 2006년도 추계학술발표회
• /
• pp.187-191
• /
• 2006

적조 종에 따라 해색이 다르다는 점에 착안해 해색 모델로부터 적조 종을 역 추정하는 기술을 개발하는데 있어 기본 입력 변수가 되는 적조 생물에 관한 광학적 특성(흡수, 역산란 특성)을 연구하였다. 실험실 수준에서의 적조 생물의 광특성을 조사하기 위해서 남해에서 채취 된 21종의 적조 생물을 배양하고 spectrophotometer를 이용해 광합성 색소에 의한 흡수계수(absorption coefficient, a)와 역산란 계수 (backscattering coefficient, $b_b$)를 측정하였다. 또한 spectrophotometer를 이용해 측정된 흡수 및 역산란 계수와 클로로필 농도를 이용하여 비 흡광계수 $(a^*)$와 비역산란계수$(b_b^*)$를 계산하였으며 이들 스펙트럼의 모양과 값을 비교하였다. 적조 생물 종에 따른 $a^*$은 파장대 440nm 에서 $0.005-0.06m^2/mg$의 값을 가지며, $b_b^*$의 범위는 $10^{-2}\sim10^{-4}m^2/mg$로 종간 약 100배의 차이가 있었다. 이와 같이 적조생물 종에 따라 스펙트럼의 모양과 값에서 차이를 나타냈으며 21종의 적조 생물 중에서 해색 스펙트럼으로 종간 구분이 가능한 종은 7-8 종이었다. 이 결과는 해색모델 개발에 있어 입력변수로 활용될 것이다.

• 한국음향학회지
• /
• 제25권5호
• /
• pp.246-256
• /
• 2006

평면파 반사 계수는 수중에서의 음파에너지에 관한 해저 바닥의 모든 정보를 담고 있고 음향 해석 모델의 입력 값으로도 사용할 수 있는 음향학적 물리량이다. 본 연구에서는 실험실 수조 환경에서 입자 매질 ( 세 종류의 유리구슬, 모래 )의 평면파 반사 계수, 음속 및 감쇠계수를 측정했다. 반사 실험은 수조의 한계를 고려해 $28{\sim}53^{\circ}$의 입사각과 중심 주파수 100kHz의 협대역 신호를 이용해 수행했다. 자기 교정법 (Self-calibration method)을 이용해 측정된 자료로부터 반사 계수를 계산했고 측정된 반사 계수의 경향 및 실험의 불확실성을 서술했다. 입자 매질의 음속 및 감쇠계수는 거리 수신 신호간의 회귀분석을 통해 계산했다. Biot 이론을 이용해 측정된 음속과 감쇠계수로부터 다공율과 침투율을 추정하고 실제 지질학적 측정값과의 유사성을 확인했다. 최종적으로 추정된 다공율, 침투율을 이용해 이론적 인 반사 계수를 계산하고 반사 실험의 측정값과 비교했다. 본 실험 결과는 Biot 이론으로 일관성 있게 입자 매질의 음향학적 물성을 설명할 수 있음을 입증한다.

• 전기학회논문지
• /
• 제67권4호
• /
• pp.569-577
• /
• 2018

In this paper, with the aid of information which is included within data, preprocessing algorithm-based black plastic classifier is designed. The slope and area of spectrum obtained by using laser induced breakdown spectroscopy(LIBS) are analyzed for each material and its ensuing information is applied as the input data of the proposed classifier. The slope is represented by the rate of change of wavelength and intensity. Also, the area is calculated by the wavelength of the spectrum peak where the material property of chemical elements such as carbon and hydrogen appears. Using informations such as slope and area, input data of the proposed classifier is constructed. In the preprocessing part of the classifier, Principal Component Analysis(PCA) and fuzzy transform are used for dimensional reduction from high dimensional input variables to low dimensional input variables. Characteristic analysis of the materials as well as the processing speed of the classifier is improved. In the condition part, FCM clustering is applied and linear function is used as connection weight in the conclusion part. By means of Particle Swarm Optimization(PSO), parameters such as the number of clusters, fuzzification coefficient and the number of input variables are optimized. To demonstrate the superiority of classification performance, classification rate is compared by using WEKA 3.8 data mining software which contains various classifiers such as Naivebayes, SVM and Multilayer perceptron.