• Speech Sciences
• /
• v.13 no.4
• /
• pp.177-186
• /
• 2006

In an effort to enhance the quality of feature vector classification and thereby reduce the recognition error rate of the speaker-independent speech recognition, we employ the Mahalanobis distance in the calculation of the similarity measure between feature vectors. It is assumed that the metric matrix of the Mahalanobis distance be diagonal for the sake of cost reduction in memory and time of calculation. We propose that the diagonal elements be given in terms of the variations of the feature vector components. Geometrically, this prescription tends to redistribute the set of data in the shape of a hypersphere in the feature vector space. The idea is applied to the speech recognition by hidden Markov model with fuzzy vector quantization. The result shows that the recognition is improved by an appropriate choice of the relevant adjustable parameter. The Viterbi score difference of the two winners in the recognition test shows that the general behavior is in accord with that of the recognition error rate.

• Proceedings of the Acoustical Society of Korea Conference
• /
• 1994.06a
• /
• pp.997-1002
• /
• 1994

Predictive neural network models are powerful speech recognition models based on a nonlinear pattern prediction. Those models can effectively normalize the temporal and spatial variability of speech signals. But those models suffer from poor discrimination between acoustically similar words. In this paper, we propose a discriminative training algorithm for predictive neural network models based on a generalized probabilistic descent (GPD) algorithm and minimum classification error formulation (MCEF). The Evaluation of our training algorithm on ten Korean digits shows its effectiveness by 40% reduction of recognition error.

• Journal of the Korean Operations Research and Management Science Society
• /
• v.36 no.3
• /
• pp.61-74
• /
• 2011

The objective of this study is to evaluate the effects of sharing uncertain yield information with a downstream supply chain player. We are interested in understanding how the amount of yield uncertainty affects the supply-side benefits and/or costs, which has not been considered in the literature, in addition to the customer-side benefits. With that purpose, this work evaluates a supplier who provides yield information in comparison with another supplier who shares no information. We simulate an order-up-to type heuristic policy that is adapted from the literature and reasonably modified to represent yield information sharing with error. From the simulation study, we argue that the customer would experience cost reduction, but the cost for supplier's inventory is increasing when sharing yield information. Furthermore, the amount of benefits and costs are situational and affected by level of yield uncertainty and demand variance. Based on the simulation study, we finally make several recommendations for the supply-side approaches to yield information sharing.

• IEIE Transactions on Smart Processing and Computing
• /
• v.2 no.4
• /
• pp.216-225
• /
• 2013

This paper discusses the issues regarding an abnormal V-shaped error confronting algebraic-based deconvolution process. Deconvolution was applied to an analysis of the effects of the Random Telegraph Noise (RTN) and Random Dopant Fluctuation (RDF) on the overall SRAM margin variations. This paper proposes a technique to suppress the problematic phenomena in the algebraic-based RDF/RTN deconvolution process. The proposed technique can reduce its relative errors by $10^{10}$ to $10^{16}$ fold, which is a sufficient reduction for avoiding the abnormal ringing errors in the RTN deconvolution process. The proposed algebraic-based analyses allowed the following: (1) detection of the truncating point of the TD-MV distributions by the screening test, and (2) predicting the MV-shift-amount by the assisted circuit schemes needed to avoid the out of specs after shipment.

• Proceedings of the Korea Concrete Institute Conference
• /
• 2000.10b
• /
• pp.1073-1078
• /
• 2000

This study is to investigate the effect of construction element such as the number of CFS ply, curing temperature, splice length and curing periods on the mechanical properties of Carbon Fiber Sheet (CFS). Through the tensile tests of CFS specimens, it can be said that the reduction factor stemmed from the number of CFS ply must be considered in the calculation of the design tensile strength of CFS. Also, the minimum splice length of CFS and curing period in $20^{\circ}C$ to satisfy the standard tension strength of CFS are over 5cm and after 3days, respectively. The measuring error of epoxy resin have no effect to tension strength of CFS until $\pm$20% error.

• The Transactions of the Korean Institute of Electrical Engineers
• /
• v.37 no.5
• /
• pp.319-326
• /
• 1988

In the presence of unmodeled dynamics, the robustness of adaptive pole assignment control using new pseudo-plant is presented. The pseudo-plant proposed by Donati et al. is modified as the gain of low pass filter can be set from zero to one. This modified pseudo-plant results in the reduction of modeling error. It is shown that not only this approach is insensitive to input frequency but also it improves the conic condition developed by Ortega et al. which is required to assure stability of adaptive control system despite the model-plant mismatch. A simple method to compensate the tracking error due to the use of pseudo-plant is considered.

• The Transactions of the Korean Institute of Electrical Engineers B
• /
• v.49 no.10
• /
• pp.708-711
• /
• 2000

Sensorless PMSM is much studied for the industrial applications and home appliances because a mechanical sensor reduce reliability and increase cost. Most of sensorless algorithms are based on motor equations, and so the magnitude of phase voltage and current should be exactly obtained. However, the inverter output voltage applied to PMSM has relatively large error in the low speed range, and a relatively poor response is shown in the low speed range. This paper investigates the improved performance of sensorless PMSM in the low speed range. This paper proposes the error reduction of inverter output voltage which is realized through the variable link voltage. The proposed algorithm is verified through simulation and experiment.

• East Asian mathematical journal
• /
• v.5 no.1
• /
• pp.95-110
• /
• 1989

In a general variance component model, nonnegative quadratic estimators of the components of variance are considered which are invariant with respect to mean value translaion and have minimum bias (analogously to estimation theory of mean value parameters). Here the minimum is taken over an appropriate cone of positive semidefinite matrices, after having made a reduction by invariance. Among these estimators, which always exist the one of minimum norm is characterized. This characterization is achieved by systems of necessary and sufficient condition, and by a cone restricted pseudoinverse. In models where the decomposing covariance matrices span a commutative quadratic subspace, a representation of the considered estimator is derived that requires merely to solve an ordinary convex quadratic optimization problem. As an example, we present the two way nested classification random model. An unbiased estimator is derived for the mean squared error of any unbiased or biased estimator that is expressible as a linear combination of independent sums of squares. Further, it is shown that, for the classical balanced variance component models, this estimator is the best invariant unbiased estimator, for the variance of the ANOVA estimator and for the mean squared error of the nonnegative minimum biased estimator. As an example, the balanced two way nested classification model with ramdom effects if considered.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.22 no.4
• /
• pp.771-781
• /
• 1997

In order to solve the problem of uncovered background region due to the region-due to the region-based motion estimation, this paper presents a new method which generates the uncovered region memory using motion estimation and shows the application of the algorithm for very low bit rate video coding. The proposed algorithm can be briefly described as follows it detects the changed region by using the information of FD(frame difference) and segmentation, and then as for only that region the backward motion estimation without transmission of shape information is done. Therefore, from only motion information the uncovered background region memory is generated and updated. The contents stored in the uncovered background region memory are referred whenever the uncovered region comes into existence. The regions with large prediction error are transformed and coded by using DCT. As results of simulation, the proposed algorithm shows the superior improvement in the subjective and objective image quality due to the remarkable reduction of transmission bits for prediction error.

• Journal of the Korean Data and Information Science Society
• /
• v.25 no.5
• /
• pp.1151-1160
• /
• 2014

In DNA microarray studies, the number of genes far exceeds the number of samples and the gene expression measures are highly correlated. Partial least squares regression (PLSR) is one of the popular methods for dimensional reduction and known to be useful for the classifications of microarray data by several studies. In this study, we suggest a modified version of the partial least squares regression to analyze gene expression data with survival information. The method is designed as a new gene selection method using PLSR with an iterative procedure of imputing censored survival time. Mean square error of prediction criterion is used to determine the dimension of the model. To visualize the data, plot for variables superimposed with samples are used. The method is applied to two microarray data sets, both containing survival time. The results show that the proposed method works well for interpreting gene expression microarray data.