• Journal of Power System Engineering
• /
• v.9 no.4
• /
• pp.58-64
• /
• 2005

주파수 영역 민감도 해석법은 동적 시스템의 전달함수에 대한 설계 파라미터의 변화에 의한 효과를 파악하기 위해 사용되어 왔으며, 이때의 민감도 함수는 시스템 설계 파라미터에 대한 시스템 전달 함수의 편미분 값이다. 일반적으로 종래의 주파수 영역 민감도 해석은 직접 미분법이나 라플라스 변환이 사용되어 왔다. 라플라스 변환을 사용하는 경우에 시스템의 차수가 증가할수록 역행렬 조작은 매우 많은 시간을 필요로 하며 또한 어려운 작업이다. 본논문에서는 이러한 다점을 보완하기 위하여 푸리에변환을 이용한 민감도 기법을 제시하였다. 파라미터의 변화에 대한 진폭-주파수 특성의 민감도 해석을 간단한 2자유도 모델과 로터 다이나믹 시스템에 적용하였다.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.14 no.4
• /
• pp.107-112
• /
• 2014

This Study is Transfer function optimization plan of volume rendering of multi user environment. Each volume data, for appropriate transfer function, they should be adjusted parameter many times. To prevent this, we propose transfer function optimization plan using crowd sourcing. In multi user environment, we use weight value for reliability level for each user. Because transfer function parameter used previous users is provided next users, they can be used effectively optimized transfer function and can reduce attempts.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2018.06a
• /
• pp.182-184
• /
• 2018

본 논문에서는 HDR 영상 신호의 고속 광전변환을 위한 파라미터 룩업 테이블 기반 구간 선형 근사 방법을 제안한다. 제안하는 방법은 고속화하기 위한 광전변환함수의 입력 값의 범위를 다수개의 구간으로 나누고 각 구간마다 별도의 선형 근사함수를 구하여 광전변환함수를 근사하고 각 구간별로 필요한 선형 근사함수의 파라미터를 룩업 테이블에 미리 저장하고 사용함으로써 보다 빠른 근사 값 계산이 가능하다. 제안한 방법의 성능 평가를 위해 MPEG 에서 제공하는 참조 소프트웨어인 HDRTools 를 기반으로 실험을 수행했고 이를 통해 참조 소프트웨어에 구현되어 있는 기존의 고속화 방법과 비교하여 더 적은 연산 수를 가지며 평균 24% 빠른 처리속도와 약 0.05dB 의 평균 PSNR 손실을 보임을 확인하였다.

• Proceedings of the KIEE Conference
• /
• 2008.07a
• /
• pp.217-218
• /
• 2008

이 논문은 이산신호에 고속푸리에 변환을 적용하여 신호에 포함되어 있는 저주파수의 진동 파라미터를 추정하는 새로운 방법에 대해서 기술하고 있다. 제안한 방법은 지수감쇠 정현파함수의 푸리에 변환에 기초를 두고 푸리에스펙트럼으로부터 직접 파라미터를 추정하는 방법이다. 푸리에스펙트럼의 첨두치와 첨두주파수 사이에 일정한 수학적 관계에서 모드를 추정하고 추정한 모드를 이용하여 모드의 크기와 위상을 추정하는 방법을 제안하고 있다. 이 논문에서 제안한 파라미터 추정방법은 수식에 기반을 둔 매우 단순한 알고리즘으로 계산속도가 매우 빠르고 작은 기억장소를 필요로 하므로 DSP 수준의 실시간 연산에 매우 적합한 알고리즘이다. 제안한 알고리즘을 간단한 시험함수에 적용한 결과, 정확하게 파라미터를 추정하여 알고리즘의 정확성을 검증하였다.

• Journal of the Korean Society for Nondestructive Testing
• /
• v.35 no.4
• /
• pp.251-257
• /
• 2015

In ultrasonic nonlinear parameter measurement using the fast Fourier transform(FFT) of tone-burst signals, the side lobe and leakage on spectrum because of finite time and non-periodicity of signals makes it difficult to measure the harmonic magnitudes accurately. The window function made it possible to resolve this problem. In this study, the effect of the Hanning and Turkey window functions on the experimental measurement of nonlinear parameters was analyzed. In addition, the effect of changes in tone burst signal number with changes in the window function on the experimental measurement was analyzed. The result for both window functions were similar and showed that they enabled reliable nonlinear parameter measurement. However, in order to restore original signal amplitude, the amplitude compensation coefficient should be considered for each window function. On a separate note, the larger number of tone bursts was advantageous for stable nonlinear parameter measurement, but this effect was more advantageous in the case of the Hanning window than the Tukey window.

• Journal of Advanced Navigation Technology
• /
• v.12 no.4
• /
• pp.378-383
• /
• 2008

This paper presents a method of image analysis based on discrete cosine transform (DCT) and fuzzy inference(Fl). It concentrated not only on the design of fuzzy inference algorithm but also on incorporating human visual parameter(HVP) into transform coefficients. In the first, HVP such as entropy, texture degree are calculated from the coefficients matrix of DCT. Secondly, using these parameters, fuzzy input variables are generated. Mamdani's operator as well as ${\alpha}$-cut function are involved to simulate the proposed approach, and consequently, experimental results are presented to testify the performance and applicability of the proposed scheme.

• KIPS Transactions on Software and Data Engineering
• /
• v.11 no.9
• /
• pp.371-380
• /
• 2022

Convolutional neural networks are widely used to manipulate data arranged in a grid, such as images. A general convolutional neural network consists of a convolutional layers and a fully connected layers, and each layer contains a nonlinear activation functions. This paper proposes a combined parametric activation function to improve the performance of convolutional neural networks. The combined parametric activation function is created by adding the parametric activation functions to which parameters that convert the scale and location of the activation function are applied. Various nonlinear intervals can be created according to parameters that convert multiple scales and locations, and parameters can be learned in the direction of minimizing the loss function calculated by the given input data. As a result of testing the performance of the convolutional neural network using the combined parametric activation function on the MNIST, Fashion MNIST, CIFAR10 and CIFAR100 classification problems, it was confirmed that it had better performance than other activation functions.

• Journal of the Korean Institute of Illuminating and Electrical Installation Engineers
• /
• v.24 no.8
• /
• pp.84-93
• /
• 2010

This paper describes a method of visualization and fast estimation of parameter in continuous time signal. The parameter estimation method of this paper directly estimate the parameters on the basis of the discrete Fourier transform. Also, this paper present to efficient visualization method of dominant parameters obtained in continuous time signal. The proposed methods are applied to test functions with three dominant modes. The results show that the proposed methods are highly applicable to parameter estimation and visualization in continuous time signal.

• Journal of Broadcast Engineering
• /
• v.16 no.6
• /
• pp.1047-1057
• /
• 2011

In this paper, we propose a novel tone mapping method that implements histogram modification framework on two local regions that are classified using K-means clustering algorithm. In addition, we propose automatic parameter tuning method for histogram modification. The proposed method enhances local details better than the global histogram method. Moreover, the proposed method is fully automatic in the sense that it does not require intervention from human to tune parameters that are involved for computing tone mapping functions. In simulations and experimental studies, the proposed method showed better performance than existing histogram modification method.

• The Journal of the Korea Contents Association
• /
• v.21 no.3
• /
• pp.616-625
• /
• 2021

Deep neural networks are an approximation method that approximates an arbitrary function to a linear model and then repeats additional approximation using a nonlinear active function. In this process, the method of evaluating the performance of approximation uses the loss function. Existing in-depth learning methods implement approximation that takes into account loss functions in the linear approximation process, but non-linear approximation phases that use active functions use non-linear transformation that is not related to reduction of loss functions of loss. This study proposes parametric activation functions that introduce scale parameters that can change the scale of activation functions and location parameters that can change the location of activation functions. By introducing parametric activation functions based on scale and location parameters, the performance of nonlinear approximation using activation functions can be improved. The scale and location parameters in each hidden layer can improve the performance of the deep neural network by determining parameters that minimize the loss function value through the learning process using the primary differential coefficient of the loss function for the parameters in the backpropagation. Through MNIST classification problems and XOR problems, parametric activation functions have been found to have superior performance over existing activation functions.