• 한국음향학회지
• /
• 제29권1호
• /
• pp.56-61
• /
• 2010

본 논문에서는 HMM 기반의 연속음성인식에서 음향모델의 최적화 기법을 논한다. 대부분의 음성인식 시스템에서 HMM 상태별로 동일한 개수의 가우시안 성분 (mixture component)을 사용해 왔다. 그러나, 음향 모델링에 사용되는 데이터 샘플의 개수는 HMM상태별로 다르므로 이에 따른 최적화를 수행할 경우 모델 파라미터의 개수를 효과적으로 줄일 수 있을 뿐 아니라, 디코딩 단계에서 음성인식기의 속도 및 인식 성능 개선이 기대된다. 본 연구에서 제안한 방법은 기존에 알려진 MDL (minimum description length) 기반의 음향모델 최적화 방법에서 가우시안 성분들의 통합과정에 가우시안 성분의 가중치 정보 (mixture weight)를 반영하도록 개선하였다. 인식 실험 결과, 제안한 방법은 가우시안 성분의 가중치를 반영하지 않는 기존 방법에 비해 향상된 최적화 성능을 보임을 확인할 수 있었다.

• 약학회지
• /
• 제13권1호
• /
• pp.47-50
• /
• 1969

The ether extract of Dang-Gui, the dried root of Angelica gigas N$_{AKAI}$, was placed into a column with aluminium oxide gel. The elution of it with a mixture of n-hexane and ethyl acetate afforded Nodakenetin (I) and the elution with a mixture of n-hexane, ethyl acetate, acetone and methyl alcohol yielded Umbelliferon (II) and Nodakenin (III). Three coumarin derivatives were identified by comparing with authentic samples. The existence of sucrose and glucose was also proved.

• ETRI Journal
• /
• 제32권5호
• /
• pp.795-800
• /
• 2010

Embedding a large vocabulary speech recognition system in mobile devices requires a reduced acoustic model obtained by eliminating redundant model parameters. In conventional optimization methods based on the minimum description length (MDL) criterion, a binary Gaussian tree is built at each state of a hidden Markov model by iteratively finding and merging similar mixture components. An optimal subset of the tree nodes is then selected to generate a downsized acoustic model. To obtain a better binary Gaussian tree by improving the process of finding the most similar Gaussian components, this paper proposes a new distance measure that exploits the difference in likelihood values for cases before and after two components are combined. The mixture weight of Gaussian components is also introduced in the component merging step. Experimental results show that the proposed method outperforms MDL-based optimization using either a Kullback-Leibler (KL) divergence or weighted KL divergence measure. The proposed method could also reduce the acoustic model size by 50% with less than a 1.5% increase in error rate compared to a baseline system.

• 품질경영학회지
• /
• 제13권2호
• /
• pp.66-72
• /
• 1985

The objective of mixture experiments with process variables is to find experimental blends and conditions that produce the product of highest quality. In this paper, designs for mixture experiments with process variables are presented, where the emphasis is on using only a fraction of the total number of possible design points and the fitting of reduced models for measuring the effects of the mixture components and process variables.

• 품질경영학회지
• /
• 제50권1호
• /
• pp.169-181
• /
• 2022

Purpose: This paper addresses a case study of developing a flame retardant wood-plastic composites (WPC) by adding tannic acid to the existing synthetic wood. The optimal mixing ratios of six components are explored to minimize the burning time using two mixture designs. Methods: In the preliminary experiment, six components are considered to find important components and their ranges. Seven D-optimal mixture design points are generated. Two points are removed for the balance of plastic components to be maintained, and the remaining five points are augmented with two basic compositions. Four components are selected to be considered in the main experiment. In the main experiment, pellets are extruded at the eight mixture design points. In-house testing of burning time is executed three times. Specimens made of pellets from two promising flame retardant compositions are sent to the accredited laboratories and tested. Results: The test results are as follows: 1) The best composition (Wood flour, Tannic acid, PE, Lubricant) = (25, 41, 10, 2) (wt%) shows the burning time of 1 second, which is 9-fold improvement compared to the the burning time of 9 seconds from the existing composition (58, 0, 10, 2) (wt%). 2) The second best composition (41, 25, 10, 2) (wt%) results in the burning time of 2 seconds. This composition is inferior to the best composition in terms of the flame retardancy, but more economical since it needs less tannic acid which is 100-fold expensive than the wood flour. Conclusion: Flame retardant compositions are found by adding tannic acid to the existing WPC employing optimal mixture designs. This case study will be helpful to practitioners who try to develop new products with additional physical properties with as small number of experimental trials as possible. Future research direction includes exploring conditions which satisfy both performance level and cost limitation simultaneously.

• 대한본초학회지
• /
• 제35권3호
• /
• pp.33-45
• /
• 2020

Objective : The present study aimed to evaluate the change of the content of 7 active components in decoctions produced by various proportional pairs of Coptis chinensis Franch and Scutellaria baicalensis Georgi in 'Mixed decoction (MD)' and 'Single decoction mixture (SDM)'. Methods : The samples of MDs were prepared by decocting C. chinensis : S. baicalensis with the ratios of 10 g:10 g, 10 g:20 g, and 20 g:10 g. Those of SDMs were prepared by blending each single decoction from C. chinensis and S. Baicalensis with the ratios of 1:1, 1:2, and 2:1. The samples were evaluated by high-performance liquid chromatography with statistical analyses. Results : The analytical methods, which were optimized and validated, were reliably applied to present research. The content of all components in both MDs and SDMs at C. chinensis : S. baicalensis = 1:1 ratio were reduced compared with single herb decoction. The components from each compositional herb in MDs were proportionally increased with the ratio of original herb increased, but inversely proportional to paired herb. The contents of components in MDs were significantly lower than those in SDMs at all ratios, except for high content of baicalin at C. chinensis : S. baicalensis = 2:1. Conclusion : It was concluded that MDs and SDMs as well as the proportions of herbs could affect the contents of the components from original herbal medicines. These results provide the information for the quality control of herbal medicine combined C. chinensis with S. baicalensis.

• 한국식품영양학회지
• /
• 제14권1호
• /
• pp.20-27
• /
• 2001

십자화과 식물인 무(Raphanus sativus L.)의 어린싹(무순)이 독특한 향과 맛으로 소비가 증가하고 있는바 그 향기성분을 SDE(simultaneous steam disti-llation/extraction) 법과 P&T(purge and cryogenic trapping)법으로 추출하여 비교, 분석하였다. 향기성분 포집시 SDE법에서는 diethyl ether 및 diethyl ether-pentane mixture(2:1, v/v)를 용매로 사용하고 P&T법에서는 diethyl ether를 사용하였다. GC와 GC-MS를 사용하여 동정한 결과 diethyl ether로 추출한 SDE법에서는 함황화합물(19종, 76.6%)이 주성분이었고, diethyl ether-pentane mixture로 추출한 경우 함황화합물(15종, 44.0%)과 탄화수소류(23종, 23.8%)가 주성분이었으며, P&T법으로 추출할 경우는 탄화수소류(25종, 84.1%)가 주성분이었다. 또한 신선한 무순의 주 향기성분은 hydrocarbon류인 n-heptane, methyl pentane이었으며, 익은 무순의 주 향기성분은 sulfur compound에 속하며 무의 주성분이기도 한 4-methylthio-3-butenyl isothiocyanate, methyl mercaptane, 2,3-dimethyl disulfide 등이었다. SDE법 보다 P&T법에 의해 추출할 경우 향기성분의 종류와 상대적인 양은 적어 분리능은 떨어지는 편이나 저분자 휘발성 물질은 오히려 더 많이 검출되어 추출방법에 따라 포집되는 향기성분의 종류와 양에 차이가 있었다.

• 대한원격탐사학회지
• /
• 제23권6호
• /
• pp.537-546
• /
• 2007

초분광영상의 분석 기법 중 하나인 선형혼합분석기법은 각 화소를 구성하는 구성물질과 구성 비율을 추정하는데 매우 유용하게 사용되고 있다. 선형혼합모델은 지질 및 광물분포와 관련된 분야에서는 비교적 성공적으로 시도되고 있으나, 산림이나 여러 인공물들로 구성된 도시와 같은 상대적으로 복잡한 구조를 가진 혼합체에서는 그 정확도가 떨어진다. 본 연구에서는 식물과 토양의 혼합체를 대상으로 선형혼합모델을 적용하여 계산된 혼합체의 반사값과 실제 이 혼합체들을 분광측정기로 측정한 반사값과의 비교를 통해, 선형혼합모델의 오차를 계산하였다. 이를 통해 선형혼합모델의 오차 원인인 구성 물질간의 분광적 상호작용이 어느 경우 발생 혹은 증가하는지를 분석하고, 또한 파장대별 상호작용의 정도 차이가 있는지를 분석하였다. 연구 결과, 선형혼합모델은 혼합체를 구성하는 구성물질의 구성비율이 비슷한 경우, 각 구성 물질간의 상호작용이 증가하여 선형혼합모델의 오차가 가장 커지는 것을 알 수 있었다. 결과적으로 선형혼합모델의 오차 원인인 구성 물질간 상호작용의 발생 정도는 혼합체를 구성하는 성분의 종류, 반사 특성, 구성비율, 파장대와 구성 성분의 배열 상태에 따라 다르게 나타나는 것을 알 수 있었다. 향후 선형혼합모델의 정확도를 높이기 위해서는 이러한 혼합체의 특징들이 구성 물질간의 상호작용에 끼치는 영향을 정량적으로 분석하여야 할 것이다.

• 산업공학
• /
• 제19권2호
• /
• pp.117-123
• /
• 2006

Many products, such as gasoline, polymer plastics, alloys, and ceramics are manufactured by mixing two or more ingredients or components. When we are to develop new mixture products, we must deal with a long list of potentially important component variables. This paper introduces some mixture screening designs, analysis tools for screening important variables, and a guideline for applying these analysis tools. The results of this paper will be helpful to engineers who work in the research and development sector of mixture product industries.

• 대한원격탐사학회지
• /
• 제25권6호
• /
• pp.547-557
• /
• 2009

LIDAR (LIght Detection And Ranging) is an active remote sensing technology which provides 3D coordinates of the Earth's surface by performing range measurements from the sensor. Early small footprint LIDAR systems recorded multiple discrete returns from the back-scattered energy. Recent advances in LIDAR hardware now make it possible to record full digital waveforms of the returned energy. LIDAR waveform decomposition involves separating the return waveform into a mixture of components which are then used to characterize the original data. The most common statistical mixture model used for this process is the Gaussian mixture. Waveform decomposition plays an important role in LIDAR waveform processing, since the resulting components are expected to represent reflection surfaces within waveform footprints. Hence the decomposition results ultimately affect the interpretation of LIDAR waveform data. Computational requirements in the waveform decomposition process result from two factors; (1) estimation of the number of components in a mixture and the resulting parameter estimates, which are inter-related and cannot be solved separately, and (2) parameter optimization does not have a closed form solution, and thus needs to be solved iteratively. The current state-of-the-art airborne LIDAR system acquires more than 50,000 waveforms per second, so decomposing the enormous number of waveforms is challenging using traditional single processor architecture. To tackle this issue, four parallel LIDAR waveform decomposition algorithms with different work load balancing schemes - (1) no weighting, (2) a decomposition results-based linear weighting, (3) a decomposition results-based squared weighting, and (4) a decomposition time-based linear weighting - were developed and tested with varying number of processors (8-256). The results were compared in terms of efficiency. Overall, the decomposition time-based linear weighting work load balancing approach yielded the best performance among four approaches.