• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.30 no.9C
• /
• pp.907-912
• /
• 2005

An efficient method for calculation of observation probability in CDHMM(Continous Density Hidden Markov Model) is proposed in this paper. the proposed algorithm, called FCOP(Fast Computation of Observation Probability), approximate obsewation probabilities in CDHMM by eliminating insignificant PDFs(Probability Density Functions) and reduces the computational load. When applied to a speech recognition system, the proposed FCOP algorithm can reduce the instruction cycles by $20\%-30\%$ and can also increase the recognition speed about $30\%$ while minimizing the loss in its recognition rate. When implemented on a practical cellular phone, the FCOP algorithm can increase its recognition speed about $30\%$ while suffering $0.2\%$ loss in recognition rate.

• Journal of the Earthquake Engineering Society of Korea
• /
• v.2 no.3
• /
• pp.83-86
• /
• 1998

The results from the Individual Plant Examination of External Event of Yonggwyang nuclear power plants, unit 3 & 4, in Korea have shown that the high degree of diversities of the experts' opinions on seismicity and attenuation models is su, pp.sed to be generic cause of uncertainty of APEs(annual exceedance probability) in the PAHA(probabilistic seismic hazard analysis). This study investigated the sensitivity of the focal depth, which is one of the most uncertain seismicity parameters in Korea, Significant differences in resultant values of annual exceedance probabilities and much more symmetrical shape of the resultant PDFs(probability density functions), in case of consideration of focal depth, are found. These two results suggest that, even for the same seismic input data set including the seismicity models and ground motion attenuation models, to consider focal depth additionally for probabilistic seismic hazard analysis evaluation makes significant influence on the distributions of uncertainties and probabilities of exceedance per year for the whole ranges of seismic hazard levels. These facts suggest that it is necessary to derive focal depth parameter more effectively from the historical and instrumental documents on earthquake phenomena in Koran Peninsula for the future study of PSHA.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.29 no.5C
• /
• pp.627-631
• /
• 2004

An efficient algorithm for clustering of GPDFs(Gaussian Probability Density Functions) in a speech recognition model is proposed in this paper. The proposed algorithm is based on CNN with the divergence as its distance measure and is applied to a speech recognition. The algorithm is compared with conventional Dk-means(Divergence-based k-means) algorithm in CDHMM(Continuous Density Hidden Markov Model). The results show that it can reduce about 31.3％ of GPDFs over Dk-means algorithm without suffering any recognition performance. When compared with the case that no clustering is employed and full GPDFs are used, the proposed algorithm can save about 61.8％ of GPDFs while preserving the recognition performance.

• International Journal of Concrete Structures and Materials
• /
• v.19 no.1E
• /
• pp.11-16
• /
• 2007

Recently, probabilistic neural network (PNN) has been proposed to predict the compressive strength of concrete for the known effect of improvement on PNN by the iteration method. However, an empirical method has been incorporated in the PNN technique to specify its smoothing parameter, which causes significant uncertainty in predicting the compressive strength of concrete. In this study, a modified probabilistic neural network (MPNN) approach is hence proposed. The global probability density function (PDF) of variables is reflected by summing the heterogeneous local PDFs which are automatically determined by the individual standard deviation of each variable. The proposed MPNN is applied to predict the compressive strength of concrete using actual test data from a concrete company. The estimated results of MPNN are compared with those of the conventional PNN. MPNN showed better results than the conventional PNN in predicting the compressive strength of concrete and provided promising results for the probabilistic approach to predict the concrete strength by using the individual standard deviation of a variable.

• Journal of Mechanical Science and Technology
• /
• v.15 no.1
• /
• pp.88-97
• /
• 2001

The present study numerically investigates the fuel-air mixing characteristics, flame structure, and pollutant emission inside a double-swirler combustor. A PSR(Perfectly Stirred Reactor) based microstructural model is employed to account for the effects of finite rate chemistry on the flame structure and NO formation. The turbulent combustion model is extended to nonadiabatic flame condition with radiation by introducing an enthalpy variable, and the radiative heat loss is calculated by a local, geometry-independent model. The effects of turbulent fluctuation are taken into account by the joint assumed PDFs. Numerical model is based on the non-orthogonal body-fitted coordinate system and the pressure/velocity coupling is handled by PISO algorithm in context with the finite volume formulation. The present PSR-based turbulent combustion model has been applied to analyze the highly intense turbulent nonpremixed flame field in the double swirler combustor. The detailed discussions were made for the flow structure, combustion effects on flow structure, flame structure, and emission characteristics in the highly intense turbulent swirling flame of the double swirler burner.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.25 no.10B
• /
• pp.1732-1743
• /
• 2000

In this paper a fixed-delay trellis search with decision feedback (FDTrS/DF) for high-density DVD systems (4.7-15GB) is proposed and implemented with FPGA. The proposed FDTrS/DF is derived by transforming the binary tree search structure into trellis search structure implying that FDTrS/DF performs better than the singnal detection techniques based on tree search structure such as FDTS/DF and SSD/DF. Advantages of FDTrS/DF are significant reductions in hardware complexity due to the unique structure of FDTrS composed of only one trellis stage requiring no traceback procedure usually implemented in the Viterbi detector. Also in this paper the PDFS/DF and SSD/DF orginally proposed for high-density magnetic recording systems are modified for the DVD system and compared with the proposed FDTrS/DF. In order to increase speed in the FPGA implementation the pipelining technique and absolute branch metric (instead of square branch metric) are applied. The proposed FDTrS/DF is shown to provide the best performance among various signal detection techniques such as PRML, DFE, FDTS/DF and SSD/DF even with a small hardware complexity.

• The Journal of the Acoustical Society of Korea
• /
• v.24 no.4
• /
• pp.224-229
• /
• 2005

In this paper, we Propose a novel cepstrum normalization method which normalizes the probability density function (pdf) of cepstrum for robust speech recognition in additive noise environments. While the conventional methods normalize the first- and/or second-order statistics such as the mean and/or variance of the cepstrum. the proposed method fully normalizes the statistics of cepstrum by making the pdfs of clean and noisy cepstrum identical to each other For the target Pdf, the generalized Gaussian distribution is selected to consider various densities. In recognition phase, we devise a table lookup method to save computational costs. From the speaker-independent isolated-word recognition experiments, we show that the Proposed method gives improved Performance compared with that of the conventional methods, especially in heavy noise environments.

• Proceedings of the Computational Structural Engineering Institute Conference
• /
• 2007.04a
• /
• pp.639-644
• /
• 2007

The accurate peak response estimation of a seismically excited structure with frictional damping system(FDS) is very difficult since the structure with FDS shows nonlinear behavior dependent on the structural period, loading characteristics, and relative magnitude between the frictional force and the excitation load. Previous studies have estimated that by replacing a nonlinear system with an equivalent linear one or by employing the response spectrum obtained based on nonlinear time history and statistical analysis. In the case that on earthquake load is defined with probabilistic characteristics, the corresponding response of the structure with FDS has probabilistic distribution. In this study, nonlinear time history analyses were performed for the structure with FDS subjected to artificial earthquake loads generated using Kanai-Tajimi filter. An equation for the probability density function (PDF) of the displacement response is proposed by adapting the PDF of the normal distribution. Finally, coefficients of the proposed PDF is obtained by regression analysis of the statistical distribution of the time history responses. Finally, the correlation between PDFs and statistical response distribution is presented.

• Journal of Korea Society of Industrial Information Systems
• /
• v.12 no.4
• /
• pp.164-168
• /
• 2007

In this paper, we propose the eye detection algorithms which can apply to the Real-Time Embedded systems. To detect the eye region, the feature vectors are obtained at the first step and then, PCA(Principal Component Analysis) and amplitude projection method is applied to composite the feature vectors. In the decision state, the estimated probability density functions (PDFs) are applied by the proposed Bayesian method to detect eye region in an image from the CCD camera. The simulation results show that our proposed method has a good detection rate on the frontal face and this can be applied to the embedded system because of its small amount of the mathematical complexity.

• Journal of the Korean Society of Safety
• /
• v.27 no.6
• /
• pp.70-77
• /
• 2012

Experimental investigations were performed to examine the characteristics of propagating flame fronts around multiple bars within a rectangular chamber. The explosion chamber is 400 mm in height, $700{\times}700mm^2$ in cross-section and has a large top-venting area, $A_v$, of $700{\times}210mm^2$. This results in a value of 0.44 for $A_v/V^{2/3}$ and a L/D value of 0.57. The multiple obstacles of length 700 mm with a blockage ratio of 30 % were placed within the chamber. Temporally resolved flame front images were recorded by a high speed video camera to investigate the interaction between the propagating flame and the obstacles. Results showed that the flame propagation speeds before the flame impinges onto the obstacle almost equal to the laminar burning velocity. As the propagating flame impinged on the obstacle, the central region of flame began to become concave, this resulted in the flame deceleration in the region. As the flame interacted with the modified flow filed generated behind the central obstacle, the probability density functions(PDFs) of the local flame displacement speed were extensively distributed toward higher speeds.