• ETRI Journal
• /
• v.38 no.3
• /
• pp.487-493
• /
• 2016

Two new methods are proposed for an unsupervised adaptation of a language model (LM) with a single sentence for automatic transcription tasks. At the training phase, training documents are clustered by a method known as Latent Dirichlet allocation (LDA), and then a domain-specific LM is trained for each cluster. At the test phase, an adapted LM is presented as a linear mixture of the now trained domain-specific LMs. Unlike previous adaptation methods, the proposed methods fully utilize a trained LDA model for the estimation of weight values, which are then to be assigned to the now trained domain-specific LMs; therefore, the clustering and weight-estimation algorithms of the trained LDA model are reliable. For the continuous speech recognition benchmark tests, the proposed methods outperform other unsupervised LM adaptation methods based on latent semantic analysis, non-negative matrix factorization, and LDA with n-gram counting.

• Proceedings of the Korean Operations and Management Science Society Conference
• /
• 2006.05a
• /
• pp.1066-1070
• /
• 2006

Independent component analysis (ICA) is a statistical method for linearly transforming observed high-dimensional multivariate data into several statistically independent components. ICA has gained wide-spread attention in a variety of fields including spectrum application. We focus on the application of ICA for separating independent sources from a set of mixtures and estimating their fractions in a mixture. The proposed method of estimating fractions is based on the regression model subject to the non-negativity constraint on coefficients. Simulation experiments are performed to demonstrate the performance of the proposed approach.

• The Korean Journal of Applied Statistics
• /
• v.16 no.2
• /
• pp.305-319
• /
• 2003

This paper provides a non-Bayesian method based on the expanded EM algorithm for segmenting the magnetic resonance images degraded by bias field. For the images with the intensity as a pixel value, many segmentation methods often fail to segment it because of the bias field(with low frequency) as well as noise(with high frequency). Our contextual approach is appropriately designed by using normal mixture model incorporated with Markov random field for noise-corrective segmentation and by using the penalized likelihood to estimate bias field for efficient bias filed-correction.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.16 no.9
• /
• pp.1870-1877
• /
• 2012

Transient signals generally have short duration and variable length with time-varying and non-stationary characteristics. Thus frame-based pattern matching method is useful for classification of transient signals. In this paper, we propose a new method for classification of underwater transient signals using a Gaussian mixture model(GMM). We carried out classification experiments for various underwater transient signals depending upon the types of noise, signal-to-noise ratio, and number of mixtures in the GMM. Experimental results have verified that the proposed method works quite well for classification of underwater transient signals.

• Macromolecular Research
• /
• v.13 no.5
• /
• pp.385-390
• /
• 2005

The electrospinning process is a fascinating method to fabricate small nanosized fibers of diameter several hundred nanometers. Surfactant-polymer solutions were prepared by adding poly(vinyl alcohol) (PVOH) to distilled water with cationic, anionic, amphoteric, and non-ionic surfactants. Average diameter of the electrospun PVOH fibers prepared from PVOH solution was over 300 nm, and was decreased to 150 nm for the mixture of PVOH/amphoteric surfactant. To explain the formation of ultra fine fiber, the characteristic properties in a mixture of PVOH/surfactant such as surface tension, viscosity, and conductivity were determined. In this paper, the effect of interactions between polymers with different classes of surfactants on the morphological and mechanical properties of electrospun PVOH nonwoven mats was broadly investigated.

• Proceedings of the Membrane Society of Korea Conference
• /
• 1995.04a
• /
• pp.34-35
• /
• 1995

The unique feature of pervaporation is the mass transfer from a liquid phase to a vapor phase through a non-porous polymeric membrane. When a liquid mixture is brought into contact with a membrane at one side, it is sorbed into the membrane. Due to a driving force applied across the membrane, the sotbed liquid molecules permeate through the membrane and evaporate at the downstream side of the membrane. In pervaporation the permeated species are usually removed from the downstream side under a relatively low vapor pressure, for example by evacuation with a vacuum pump. As far as this condition is fulfilled, the evaporation step can be considered to be much faster than sorption or diffusion. Hence evaporation does not contribute to permselectivity. Therefore the separation by pervaporation results from the differences in the preferential sorption of the individual components of a mixture into the membrane together with the diffusion rates through the membrane. This postulation implies that both sorption and diffusion phenomena have to be accounted for to understand the physico-chemical nature of the pervaporation separation process.

• Proceedings of the Korean Society of Agricultural Engineers Conference
• /
• 1999.10c
• /
• pp.301-306
• /
• 1999

Over 4.1 billion metric tons of non-hazardous solid wastes are generated in the United States anually. Therefore, there is a need to reuse or reprocess as many products or by -products as possible. In this report, the aggregates produced from Naval Shipyard and Charlexton area dredge spoils are designated as NC and DS, respectively. The major objective of this portion was to determine the feasibility of the use of these aggregates in concrete mixtures. According to test results, the unit weight of concrete mixtures containing NC and DS (vitrified glass) was lower than the control mixture, and the control mixture produced higher compressive and bending strength compared to mixtures containing vitrified glass. However the mixtures containing vitrified glass still could be used in the field based on the minimum criteria set for the mix desgin.

• Korean Journal of Air-Conditioning and Refrigeration Engineering
• /
• v.12 no.12
• /
• pp.1049-1056
• /
• 2000

In this study, condensation heat transfer coefficients(HTCs) of 2 nonazeotropic refrigerant mixtures of HFC32/HFC134a and HFC134a/HCFC123 at various compositions were measured on a horizontal smooth tube. All data were taken at the vapor temperature of 39$^{\circ}C$ with a wall subcooling of 3~8K. Test results showed that HTCs of tested mixtures were 11.0~85.0% lower than the ideal values calculated by the mass fraction weighting of the pure components HTCs. Thermal resistance due to the diffusion vapor film was partly responsible for the significant reduction of HTCs with these nonazeotropic mixtures. The measured data were compared against the predicted ones by Colburn and Drew\`s film model and a good agreement was observed.

• Journal of Nutrition and Health
• /
• v.9 no.3
• /
• pp.26-30
• /
• 1976

The contents of sterols in bean sprouts have been determined by mass spectrical analysis. From non-saponifiable fraction of n-Hexane extracts of the under ground parts sterol-like substance $(m.P.\;142-145^{\circ}C)$ was isolated. It was shown to be a sterol mixture on the basis of three peaks in gas liquid chromatogram. By mass spectrometrical analysis of the mixture, it was composed of 4.52mg% of Campesterol $(C_{28},\;{\Delta}^{5})$, 10.35mg% of ${\beta}-sitosterol\;(C_{29},\;{\Delta}^{5})$ and 7.75mg% of stigmasterol $(C_{29},\;{\Delta}^{5,\;22})$.

• Bulletin of the Korean Chemical Society
• /
• v.15 no.3
• /
• pp.214-221
• /
• 1994

We present some results obtained from theoretical study on a non-symmetrical A/BC polymeric system including solvent which consists of two phases, a polymeric phase A on one side and a mixture of polymers B (as a compatibilizer) and C on the other in the presence of a solvent. By employing the functional integral techniques we derive the mean-field equations and solve them numerically to deduce the physical properties of the interface involving the polymers and solvent concentration profiles in the limit that molecular weights of all the polymers involved tend to infinity. The calculations are performed for typical values of the Flory interaction parameters and for the volume fraction of polymer B in the asymptotic phase and of solvent. In the polymers/solvent blend under consideration the interfacial adsorption of polymer B, the solvent concentration, and degrees of the specific interaction between the polymers are found to play important roles in modification of the interfacial properties.