• Proceedings of the Acoustical Society of Korea Conference
• /
• spring
• /
• pp.69-72
• /
• 2002

본 논문에서 화자공간모델 evolution에 기반한 continuous density hidden Markov model (CDHMM)의 online 적응에 대한 새로운 기법을 제안한다. 학습화자의 a priori knowledge을 나타내는 화자공간모델은 factor analysis (FA) 또는 probabilistic principal component analysis (PPCA)와 같은 은닉변수모델(latent variable model)에 의해 효과적으로 나타내어진다. 은닉 변수모델은 화자공간모델뿐아니라 CDHMM 파라메터의 ajoint prior분포를 표시함으로, maximum a posteriori(MAP)적응기법에 직접 적용되어진다. 화자공간모델의 hyperparameters와 CDHMM파라메터를 동시에 순차적으로 적응하기 위해 quasi-Bayes (QB)추정 기술에 기반한 online 적응기법을 제안한다. 연속숫자음 인식과 관련된 화자적응 실험을 통해 제안된 기법은 적은 적응데이터에서 좋은 성능을 나타내며, 데이터가 증가함에 따라 성능이 지속적으로 증가함을 보여준다.

• Journal of Korean Society for Quality Management
• /
• v.47 no.2
• /
• pp.255-269
• /
• 2019

Purpose: We examine the impact of Customization of TV Home Shopping Mobile Applications on Service Quality and Customer Loyalty. Methods: We collect data by using survey and proposed relationships between latent variables using Structural Equations Modeling. Results: We find that Customization of TV Home Shopping Mobile Applications positively affect Service Quality and Customer Loyalty. In addition, service quality work as a mediate variable between Customization with Customer Loyalty. Conclusion: Our findings explain the significant relationships between each variables and hence it is necessary to consider Customization and Service Quality in order to achieve desired goal of enhancing business performance through mobile applications.

• Communications for Statistical Applications and Methods
• /
• v.29 no.1
• /
• pp.27-40
• /
• 2022

In this paper, the focus on the removal noise in the binary image based on the variational Bayesian method with the Ising model. The observation and the latent variable are the degraded image and the original image, respectively. The posterior distribution is built using the Markov random field and the Ising model. Estimating the posterior distribution is the same as reconstructing a degraded image. MCMC and variational Bayesian inference are two methods for estimating the posterior distribution. However, for the sake of computing efficiency, we adapt the variational technique. When the image is restored, the iterative method is used to solve the recursive problem. Since there are three model parameters in this paper, restoration is implemented using the VECM algorithm to find appropriate parameters in the current state. Finally, the restoration results are shown which have maximum peak signal-to-noise ratio (PSNR) and evidence lower bound (ELBO).

• Proceedings of the Korean Institute of Building Construction Conference
• /
• 2022.04a
• /
• pp.174-175
• /
• 2022

The blast furnace slag, which is widely used as a cement admixture, has latent hydraulics under the influence of cement hydrate, and fly ash and silica fume mainly cause a pozolane reaction. As a result, the cement structure becomes dense, and it is possible to compensate for defects when concrete is usually made with portland cement alone. When fixing carbon dioxide through reaction with carbon dioxide, the amount of calcium hydroxide in the cement paste is important. The larger the amount of calcium hydroxide, the more active the reaction may occur. It is also an important variable in calculating the depth of neutralization through carbonization. In this study, calcium hydroxide in cement paste using mixed materials was quantified through thermal analysis.

• Journal of the Korean Dietetic Association
• /
• v.25 no.2
• /
• pp.142-155
• /
• 2019

The aim of this study was to construct and test a structural equation model for the risk factors of metabolic syndrome in Korean adults. The structural equation model hypothesizes that eating alone and feeling depressed is a risk factor for metabolic syndrome. The data of this study were obtained from the Sixth Korea National Health and Nutrition Examination Survey which was cross-sectional data from the representative national survey. A total of 4,013 subjects replied to the survey item of lifestyle and completed the physical examinations among adults aged 19 years or older in South Korea was in 2015. The structural model in this study was composed of four latent variables: eating alone, depression, negative health behavior, and metabolic syndrome. Two variables, the rate of eating alone and depression, were exogenous variables. Negative health behavior was both a mediating variable and endogenous variable, and metabolic syndrome was the final endogenous variable. The data were analyzed using the Maximum Likelihood method and bootstrapping. The structural model was appropriate for the data based on the model fit indices. The results of this study can be summarized as follows: Eating alone is a direct risk factor of metabolic syndrome in Korean women. Depression can mediate metabolic syndrome through negative health behaviors. Negative health behavior had a direct impact on metabolic syndrome in both men and women. This study may be a guideline for interventions and strategies to reduce the incidence of metabolic syndrome in Korean adults.

• The Korean Journal of Applied Statistics
• /
• v.35 no.2
• /
• pp.217-227
• /
• 2022

Quantile regression is widely used in many fields based on the advantage of providing an efficient tool for examining complex information latent in variables. However, modern large-scale and high-dimensional data makes it very difficult to estimate the quantile regression model due to limitations in terms of computation time and storage space. Divide-and-conquer is a technique that divide the entire data into several sub-datasets that are easy to calculate and then reconstruct the estimates of the entire data using only the summary statistics in each sub-datasets. In this paper, we studied on a variable selection method using Bayes information criteria by applying the divide-and-conquer technique to the penalized quantile regression. When the number of sub-datasets is properly selected, the proposed method is efficient in terms of computational speed, providing consistent results in terms of variable selection as long as classical quantile regression estimates calculated with the entire data. The advantages of the proposed method were confirmed through simulation data and real data analysis.

• Journal of the Korean Society of Costume
• /
• v.64 no.7
• /
• pp.57-67
• /
• 2014

This study aims to provide visual data from analysis of the Depaysement approaches with new viewpoints to inspire and develop new fashion window design ideas. The literature and existing researches related to Depaysement were analyzed for theoretical review, and Depaysement expression approaches were identified by expression characteristics. Theme concepts using traditional Korean images, which could be applied to fashion window displays in Korea, were established, and K(Korean)-fashion design was created to develop fashion window display design. Then, the Depaysement fashion window display was executed using Adobe Illustrator and Photoshop. The results of this study are summarized below. 'Change of forms and materials' could visualize the factors inducing curiosity, which can directly stimulate the consumption sentiment lying latent in the mind of observers by assigning new values to fashion goods displayed inside windows. Unconscious experience and remarkable stories, which are not possible to encounter in an everyday setting, can be visualized through the window display in 'heterogeneous combination of objects.' 'The location change of an object' could express the refreshing and shocking scene to give weird anxiety and mental contradiction to observers by fashion window display, which could break fixed idea of human beings. 'The change of object awareness' could express contradiction and denial, which could liberate the unconsciousness lying latent inside observers through fashion window display. 'Change of spatial awareness' could create the design which maximized the fashion images of goods displayed by helping the observers to change the space of their unconsciousness selectively at their will through the fashion window display with hidden, strange, ambiguous and variable image like a riddle.

• Journal of KIISE:Software and Applications
• /
• v.32 no.11
• /
• pp.1071-1083
• /
• 2005

By estimating probability distributions of the good solutions in the current population, some researchers try to find the optimal solution more efficiently. Particularly, finite mixtures of distributions have a very useful role in dealing with complex problems. However, it is difficult to choose the number of components in the mixture models and merge superior partial solutions represented by each component. In this paper, we propose a new continuous evolutionary optimization algorithm with distribution estimation by variational Bayesian mixtures of factor analyzers. This technique can estimate the number of mixtures automatically and combine good sub-solutions by sampling new individuals with the latent variables. In a comparison with two probabilistic model-based evolutionary algorithms, the proposed scheme achieves superior performance on the traditional benchmark function optimization. We also successfully estimate the parameters of S-system for the dynamic modeling of biochemical networks.

• Phonetics and Speech Sciences
• /
• v.11 no.3
• /
• pp.39-47
• /
• 2019

We propose a method to improve the performance of spontaneous speech recognizers by extending their phone set using speech data. In the proposed method, we first extract variable-length phoneme-level segments from broadcast speech signals, and convert them to fixed-length latent vectors using an long short-term memory (LSTM) classifier. We then cluster acoustically similar latent vectors and build a new phone set by choosing the number of clusters with the lowest Davies-Bouldin index. We also update the lexicon of the speech recognizer by choosing the pronunciation sequence of each word with the highest conditional probability. In order to analyze the acoustic characteristics of the new phone set, we visualize its spectral patterns and segment duration. Through speech recognition experiments using a larger training data set than our own previous work, we confirm that the new phone set yields better performance than the conventional phoneme-based and grapheme-based units in both spontaneous speech recognition and read speech recognition.

• Journal of Broadcast Engineering
• /
• v.24 no.3
• /
• pp.387-394
• /
• 2019

In this study, a new method of audio high-band coding based on autoencoder with side information is proposed. The proposed method operates in the MDCT domain, and improves the performance by using additional side information consisting of the previous and current low bands, which is different from the conventional autoencoder that only inputs information to be encoded. Moreover, the side information in a time-frequency domain enables the high-band coder to utilize temporal characteristics of the signal. In the proposed method, the encoder transmits a 4-dimensional latent vector computed by the autoencoder and a gain variable using 12 bits for each frame. The decoder reconstructs the high band by applying the decoded low bands in the previous and current frames and the transmitted information to the autoencoder. Subjective evaluation confirms that the proposed method provides equivalent performance to the SBR at approximately half the bit rate of the SBR.