• Proceedings of the KSPS conference
• /
• 2003.05a
• /
• pp.123-126
• /
• 2003

We discuss how to reduce the number of inverse matrix and its dimensions requested in MLLR framework for speaker adaptation. To find a smaller set of variables with less redundancy, we employ PCA(principal component analysis) and ICA(independent component analysis) that would give as good a representation as possible. The amount of additional computation when PCA or ICA is applied is as small as it can be disregarded. The dimension of HMM parameters is reduced to about 1/3 ~ 2/7 dimensions of SI(speaker independent) model parameter with which speech recognition system represents word recognition rate as much as ordinary MLLR framework. If dimension of SI model parameter is n, the amount of computation of inverse matrix in MLLR is proportioned to O($n^4$). So, compared with ordinary MLLR, the amount of total computation requested in speaker adaptation is reduced to about 1/80~1/150.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.5 no.1
• /
• pp.228-246
• /
• 2011

A 3D model based approach for a face representation and recognition algorithm has been investigated as a robust solution for pose and illumination variation. Since a generative 3D face model consists of a large number of vertices, a 3D model based face recognition system is generally inefficient in computation time and complexity. In this paper, we propose a novel 3D face representation algorithm based on a pixel to vertex map (PVM) to optimize the number of vertices. We explore shape and texture coefficient vectors of the 3D model by fitting it to an input face using inverse compositional image alignment (ICIA) to evaluate face recognition performance. Experimental results show that the proposed face representation and recognition algorithm is efficient in computation time while maintaining reasonable accuracy.

• Proceedings of the KIEE Conference
• /
• 1995.07a
• /
• pp.6-8
• /
• 1995

In this paper, we study the computation of the electric field of dielectric analysis models with the conductivity on its surface. The finite element formulation describes a sinusoidal electrodynamic field computation. One term is added to this functional in order to take the conductivity on its surface into accounts. The electric field computations of the dielectric analysis model are done first with the surface conductivity and second with the volume conductivity. Also, it is shown that a surface conductor with sufficiently large conductivity can be substituted with a floating equipotential line. This method is applied to an insulator in arbitrary shape with the conductivity on its surface.

• Journal of the Korean Society for Industrial and Applied Mathematics
• /
• v.12 no.2
• /
• pp.119-131
• /
• 2008

The paper considers the computation method in the performance evaluation of cellular network in the phase-type distribution assumptions that the channel holding times induced from mobility are modeled by well-fitted distributions to reflect an actual situation. When ww consider a phase-type distribution model instead of exponential distribution, the complexity of the computation increase exponential even though the accuracy is improved. We consider an efficient numerical algorithm to compute the performance evaluations in cellular networks such as a handoff call dropping probability, new call blocking probability, and handoff arrival rate. Numerical experiment shows that numerical analysis results are well approximated to the results of simulation.

• Journal of Ecology and Environment
• /
• v.33 no.4
• /
• pp.275-288
• /
• 2010

Ecological modeling faces some unique problems in dealing with complex environment-organism relationships, making it one of the toughest domains that might be encountered by a modeler. Newer technologies and ecosystem modeling paradigms have recently been proposed, all as part of a broader effort to reduce the uncertainty in models arising from qualitative and quantitative imperfections in the ecological data. In this paper, evolutionary computation modeling approaches are introduced and proposed as useful modeling tools for ecosystems. The results of our case study support the applicability of an algal predictive model constructed via genetic programming. In conclusion, we propose that evolutionary computation may constitute a powerful tool for the modeling of highly complex objects, such as river ecosystems.

• Journal of international Conference on Electrical Machines and Systems
• /
• v.1 no.2
• /
• pp.10-17
• /
• 2012

This paper presents an improved iron loss model, for the computation of the no load iron loss in the stator core of the in-wheel permanent magnet synchronous motors (PMSM), for the cases of with and without stator skew. 2-D analytical model is used for the computation of tooth and yoke flux densities of the in-wheel PMSM. The no load iron loss computed by the improved iron loss model, for the cases of with and without skew is compared with the finite element method (FEM) and the results show good consistency.

• Proceedings of the KIEE Conference
• /
• 2007.07a
• /
• pp.1132-1133
• /
• 2007

A circuit-domain model of PWM inverter provides accurate simulation results in consideration of detail switching characteristics. Although, a huge amount of computation time is demanded for the simulation results of several ten seconds, which is the required time to analyze system behaviors or control performances of Electric Power Steering(EPS) on real drive condition. This paper describes the nonlinear inverter model for fast dynamic simulation of EPS without the PWM concept through analyzing the effect of nonlinear switching characteristics like dead time, forward voltage drop and conduction resistance. Some inverter models including proposed model are compared from two standpoints which are computation time and accuracy. The comparison results show the usefulness of the developed model in order to develop the control algorithm through the fast prediction of system behaviors.

• Journal of the Korean Data and Information Science Society
• /
• v.14 no.1
• /
• pp.1-9
• /
• 2003

In this paper we discuss Bayesian computation and model selection for Littlewood-Verrall model using Gibbs sampling. A numerical example with a simulated data is given.

• Computers and Concrete
• /
• v.7 no.4
• /
• pp.303-315
• /
• 2010

In this paper, we present a new finite Timoshenko beam element with a model for ultimate load computation of reinforced concrete frames. The proposed model combines the descriptions of the diffuse plastic failure in the beam-column followed by the creation of plastic hinges due to the failure or collapse of the concrete and or the re-bars. A modified multi-scale analysis is performed in order to identify the parameters for stress-resultant-based macro model, which is used to described the behavior of the Timoshenko beam element. The micro-scale is described by using the multi-fiber elements with embedded strain discontinuities in mode 1, which would typically be triggered by bending failure mode. A special attention is paid to the influence of the axial force on the bending moment - rotation response, especially for the columns behavior computation.

• The KIPS Transactions:PartA
• /
• v.18A no.4
• /
• pp.143-150
• /
• 2011

This paper presents an EDF based scheduling algorithm for scheduling imprecise computation model where each task consists of mandatory part and optional part. Imprecise computation is useful to manage overload condition. In overload situation, some optional parts should be removed. The proposed DOP algorithm removes optional parts of earlier deadline tasks to enhance flexibly for newly arriving tasks. A simulation result shows that DOP has better performance than other algorithms.