• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.38C no.6
• /
• pp.503-511
• /
• 2013

This paper presents a speech enhancement method using non-negative matrix factorization. In training phase, we can obtain each basis matrix from speech and specific noise database. After training phase, the noisy signal is separated from the speech and noise estimate using basis matrix in enhancement phase. In order to improve the performance, we model the change of encoding matrix from training phase to enhancement phase using independent Gaussian distribution models, and then use the constraint of the objective function almost same as that of the above Gaussian models. Also, we perform a smoothing operation to the encoding matrix by taking into account previous value. Last, we apply the Log-Spectral Amplitude type algorithm as gain function.

• Journal of the Institute of Electronics Engineers of Korea SD
• /
• v.41 no.2
• /
• pp.9-15
• /
• 2004

We present a new unified analytical front surface potential model, which can accurately describe the transitions between the partially-depleted (PD) and the fully-depleted (FD) regimes with an analytical expression for the critical voltage V$_{c}$ delineating the PD and the FD region. It is valid in all regions of operation (from the sub -threshold to the strong inversion) and has the shorter calculation time than the iterative procedure approach. A charge sheet model based on the above explicit surface potential formulation is used to derive a single formula for the drain current valid in all regions of operation. Most of the secondary effects can be easily included in the charge sheet model and the model accurately reproduces various numerical and experimental results. No discontinuity in the derivative of the surface potential is found even though three types of smoothing functions are used. More importantly, the newly introduced parameters used in the smoothing functions do not strongly depend on the process parameter.

• Journal of the Korean Data and Information Science Society
• /
• v.27 no.4
• /
• pp.993-1000
• /
• 2016

Semi-supervised learning makes it easy to use an unlabeled data in the supervised learning such as classification. Applying the semi-supervised learning on the regression analysis, we propose two methods for a better regression function estimation. The proposed methods have been assumed different marginal densities of independent variables and different smoothing parameters in unlabeled and labeled data. We shows that the overfitted pilot estimator should be used to achieve the fastest convergence rate and unlabeled data may help to improve the convergence rate with well estimated smoothing parameters. We also find the conditions of smoothing parameters to achieve optimal convergence rate.

• Journal of Korean Society for Quality Management
• /
• v.32 no.4
• /
• pp.265-273
• /
• 2004

We got the growth distance curve by spline smoothing method with observed marketing data and the growth velocity curve by the derivation of the growth distance curve. Using this growth velocity curve, we defined the several characteristic points which describe the variation of marketing data. In this paper, to specify several patterns of marketing data, we suggested characteristic function by using these characteristic points. In addition, we applied characteristic function to the seventeen brands of electric home products data.

• Proceedings of the KSR Conference
• /
• 2000.05a
• /
• pp.151-156
• /
• 2000

The brake system is important to stop train safely. The train is sloped by regenerative brake and pneumatic brake which are continuously blended at service brake. When service is applied to train, it is controlled by train weight and brake command. The jerk limitation function is applied for impulseless smoothing braking. All brake applications in service condition have a function of the variable load control to keep the braking effort in proportion to each car load. All of control function are performed by brake controller. Therefore, we will propose the design of brake control system in order to control efficiently

• International Journal of Automotive Technology
• /
• v.8 no.3
• /
• pp.337-342
• /
• 2007

A multi-objective optimization technique was implemented to obtain optimal topologies of the inner reinforcement for a vehicle's hood simultaneously considering the static stiffness of bending and torsion and natural frequency. In addition, a smoothing scheme was used to suppress the checkerboard patterns in the ESO method. Two models with different curvature were chosen in order to investigate the effect of curvature on the static stiffness and natural frequency of the inner reinforcement. A scale factor was employed to properly reflect the effect of each objective function. From several combinations of weighting factors, a Pareto-optimal topology solution was obtained. As the weighting factor for the elastic strain efficiency went from 1 to 0, the optimal topologies transmitted from the optimal topology of a static stiffness problem to that of a natural frequency problem. It was also found that the higher curvature model had a larger static stiffness and natural frequency than the lower curvature model. From the results, it is concluded that the ESO method with a smoothing scheme was effectively applied to topology optimization of the inner reinforcement of a vehicle's hood.

• The Proceedings of the Korean Institute of Illuminating and Electrical Installation Engineers
• /
• v.6 no.3
• /
• pp.43-49
• /
• 1992

Forecasting of electric power demand has been a basic element for electric power system operation and system development, and it's accuracy has very strong influence on reliability and economical efficience of power supply. So, in this paper, long―term maximum electric power demand has been forecasted by using the triple exponential smoothing method initiated R.G.Brown. It has been regarded this method as high accuracy and operational convenience. The smoothing function is a liner combination of all past observations and the weight given to previous observations decreases geometrically with age.

• Journal of the Institute of Convergence Signal Processing
• /
• v.4 no.4
• /
• pp.47-51
• /
• 2003

Due to the asymptotic property, deadbeat control which is well used in digital control system can not be applied to the continuous time system. But recently by use of the finite Laplace Transform to transfer function and establishment of some settling conditions, CDBC(Continuous time Deadbeat Control) is studied. For CDBC design, transfer function is constituted with delay elements and then order and interpolation conditions are derived. In other way, digital deadbeat controller is implemented and it's output is changed to continuous type by smoothing elements. In this paper multirate sampling is used and so inner controller is sampled faster than output feedback loop. And End order smoothing elements is placed to the output of digital deadbeat controller. By the multirate sampling overall output response is improved. The controller is impleneted as a serial integral compensator in the forward path and a local feedback compensator introduced into the outpute feedback loop. Matlab Simulink is used for simulation.

• Journal of KSNVE
• /
• v.7 no.2
• /
• pp.319-329
• /
• 1997

Wigner-Ville distribution has been recognized as a useful tool and applied to various types of mechanical noise and vibration signals, but its limitation which mainly comes from the cross-talk has not been well addressed. The cross-talk takes place for a signal with multiple components, simply because the Wigner-Ville distribution is a bilinear transform. The cross-talk often causes a negative value in the distribution. This cannot be accepted for the Wigner- Ville distribution, because it is an expression of power. Smoothing the Wigner-Ville distribution by convoluting it wih a window, is most commonly used to reduce the cross-talk. There can be infinite number of distributions depending on the windows. In this paper, we attempted to develop a distribution which is the best or the optimal in reducing the cross-talk. This could be possible by employing the ambiguity function. For a general signal, however it is difficult to express the ambiguity function as a mathematically closed form. This requires an appropriate modeling to make such expression possible. We approximated the Wigner-Ville distribution as a sum of linear segments. In the ambiguity function domain, the legitimate components are reflected as linear lines passing through the origin. Every lines has its own length and slope. But, the cross-talk is widely distributed in the ambiguity function plane. Based on this realization, we proposed a two-dimensional window which is in fact 'rotating window', that can eliminate cross-talk component. The rotating window is examined numerically and is found to have a better performance in reducing the cross-talk than conventional windows, the Gaussian window.

• Communications for Statistical Applications and Methods
• /
• v.21 no.1
• /
• pp.105-114
• /
• 2014

In this paper we suggest an efficient method to estimate the distribution function using the Bezier curve, and compare it with existing methods by simulation studies. In addition, we suggest a robust version of cross-validation criterion to estimate the number of Bezier points, and showed that the proposed method is better than the existing methods based on simulation studies.