• MALSORI
• /
• no.49
• /
• pp.107-121
• /
• 2004

In this paper, we describe a cross-morpheme pronunciation variation model which is especially useful for constructing morpheme-based pronunciation lexicon to improve the performance of a Korean LVCSR. There are a lot of pronunciation variations occurring at morpheme boundaries in continuous speech. Since phonemic context together with morphological category and morpheme boundary information affect Korean pronunciation variations, we have distinguished phonological rules that can be applied to phonemes in within-morpheme and cross-morpheme. The results of 33K-morpheme Korean CSR experiments show that an absolute reduction of 1.45% in WER from the baseline performance of 18.42% WER was achieved by modeling proposed pronunciation variations with a possible multiple context-dependent pronunciation lexicon.

• MALSORI
• /
• no.66
• /
• pp.105-116
• /
• 2008

In this research, we propose a speaker identification system using a kernel method which is expected to model the non-linearity of speech features well. We have been using principal component analysis (PCA) successfully, and extended to kernel PCA, which is used for many pattern recognition tasks such as face recognition. However, we cannot use kernel PCA for speaker identification directly because the storage required for the kernel matrix grows quadratically, and the computational cost grows linearly (computing eigenvector of $l{\times}l$ matrix) with the number of training vectors I. Therefore, we use greedy kernel PCA which can approximate kernel PCA with small representation error. In the experiments, we compare the accuracy of the greedy kernel PCA with the baseline Gaussian mixture models using MFCCs and PCA. As the results with limited enrollment data show, the greedy kernel PCA outperforms conventional methods.

• Structural Engineering and Mechanics
• /
• v.23 no.3
• /
• pp.233-247
• /
• 2006

Most codes of practice state that for large in-plane structures it is necessary to account for the spatial variability of earthquake ground motion. There are essentially three effects that contribute for this variation: (i) wave passage effect, due to finite propagation velocity; (ii) incoherence effect, due to differences in superposition of waves; and (iii) the local site amplification due to spatial variation in geological conditions. This paper discusses the procedures to be undertaken in the time domain analysis of a cable-stayed bridge under spatial variability of earthquake ground motion. The artificial synthesis of correlated displacements series that simulate the earthquake load is discussed first. Next, it is described the 3D model of the International Guadiana Bridge used for running tests with seismic analysis. A comparison of the effects produced by seismic waves with different apparent propagation velocities and different geological conditions is undertaken. The results in this study show that the differences between the analysis with and without spatial variability of earthquake ground motion can be important for some displacements and internal forces, especially those influenced by symmetric modes.

• Journal of the Korean Data and Information Science Society
• /
• v.18 no.2
• /
• pp.553-560
• /
• 2007

In this paper, we consider two components system which lifetimes have Freund's bivariate exponential model with equal failure rates. We propose Bayesian multiple comparisons procedure for the failure rates of I Freund's bivariate exponential populations based on Dirichlet process priors(DPP). The family of DPP is applied in the form of baseline prior and likelihood combination to provide the comparisons. Computation of the posterior probabilities of all possible hypotheses are carried out through Markov Chain Monte Carlo(MCMC) method, namely, Gibbs sampling, due to the intractability of analytic evaluation. The whole process of multiple comparisons problem for the failure rates of bivariate exponential populations is illustrated through a numerical example.

• Proceedings of the KIEE Conference
• /
• summer
• /
• pp.639-641
• /
• 2004

This paper presents the evaluaton procedures and the estimation model for DSM Potential by stage on residential sector in Korea. In general, the evaluation process of the potential savings for DSM measures or programs consists of baseline electricity consumption forecast and potential evaluation such as maximum technical potential(MTP), phased-in technical potential(PTP), economic potential (EP), and achievable potential(AP). The purpose of this paper is to establish the evaluation process of those DSM potential for residential sector.

• The KSFM Journal of Fluid Machinery
• /
• v.13 no.2
• /
• pp.12-17
• /
• 2010

A kind of disk-shape boundary layer pump is designed numerically by using a software of computational fluid dynamics, which is widely used for the special purposes such as artificial hearts, bio-fluidics and transportation of oceanic lives, etc. From the numerical simulation with an axisymmetric model, some benchmark problems are tested and compared with experimental results. The performance of disk pump is graphically visualized from the computational results, and converted to the dimensionless parameters. Finally, the obtained numerical data in the present investigation can be used for the baseline for new design to achieve a more efficient disk pump.

• MALSORI
• /
• no.67
• /
• pp.149-166
• /
• 2008

In spoken document retrievals (SDR), subword (typically phonemes) indexing term is used to avoid the out-of-vocabulary (OOV) problem. It makes the indexing and retrieval process independent from any vocabulary. It also requires a small corpus to train the acoustic model. However, subword indexing term approach has a major drawback. It shows higher word error rates than the large vocabulary continuous speech recognition (LVCSR) system. In this paper, we propose an probabilistic slot detection and n-gram based string matching method for phone based spoken document retrievals to overcome high error rates of phone recognizer. Experimental results have shown 9.25% relative improvement in the mean average precision (mAP) with 1.7 times speed up in comparison with the baseline system.

• Transactions of the Korean Society of Mechanical Engineers A
• /
• v.33 no.1
• /
• pp.42-48
• /
• 2009

A design methodology for control strategy and control structure gives a direct impact on wind turbine's performance and life cycle. A baseline control law which is a variable rotor speed and variable pitch control strategy is introduced, and a mathematic performance model of a wind turbine dynamics is derived. By using a numeric optimization algorithm, the steady state operating conditions of wind turbines are identified. Because aerodynamic interaction of winds with rotor blades is basically nonlinear, a linearization procedure is applied to analyze wind turbine dynamic variations for whole operating conditions. It turns out the wind turbine dynamics vary much depending on its operating condition.

• New & Renewable Energy
• /
• v.6 no.3
• /
• pp.47-54
• /
• 2010

Foundation plays an important role in the offshore wind turbine system. Different from conventional foundations, the suction caisson is proven to be economical and reliable. In this work, three-dimensional finite element method is used to check the suitability of suction caisson foundation. NREL 5MW wind turbine is chosen as a baseline model in our simulation. The maximum overturning moment and vertical load at the mudline are calculated using FAST and Bladed. Meanwhile the soil-structure interaction response from our simulation is also compared with the experiment data from Oxford university. The design parameter such as caisson length, diameter of skirt and spacing of multipod are investigated. Accordingly based on these parameters suggestions are given to use suction caisson foundations more efficiently.

• Journal of the Korean Institute of Telematics and Electronics B
• /
• v.32B no.5
• /
• pp.765-776
• /
• 1995

This paper describes a speaker adaptation method to improve the recognition performance of MLP(multiLayer Perceptron) based HMM(Hidden Markov Model) speech recognizer. In this method, we use lst-order linear transformation network to fit data of a new speaker to the MLP. Transformation parameters are adjusted by back-propagating classification error to the transformation network while leaving the MLP classifier fixed. The recognition system is based on semicontinuous HMM's which use the MLP as a fuzzy vector quantizer. The experimental results show that rapid speaker adaptation resulting in high recognition performance can be accomplished by this method. Namely, for supervised adaptation, the error rate is signifecantly reduced from 9.2% for the baseline system to 5.6% after speaker adaptation. And for unsupervised adaptation, the error rate is reduced to 5.1%, without any information from new speakers.