• Proceedings of the IEEK Conference
• /
• 2003.07e
• /
• pp.2204-2207
• /
• 2003

본 논문은 LMS 적응 필터의 수렴성능의 향상을 위한 새로운 적응신호처리 기법을 제안한다. 기존의 LMS 알고리즘의 문제점으로는 수렴특성이 입력 벡터의 자기 상관행렬의 고유치 분포에 영향을 받는다는 점이다. 본 논문은 두 선행처리 기법, 즉, 직교 변환에 의한 선행처리 기법과 split filter 구조 필터링 기법을 결합하여 보다 개선된 수렴특성을 갖는 적응신호처리 기법을 제안한다.

• Investigative Magnetic Resonance Imaging
• /
• v.6 no.2
• /
• pp.137-146
• /
• 2002

Purpose : This paper introduces an improved classification approach which adopts a deterministic relaxation method and an agglomerative clustering technique for the classification of MRI using neural network. The proposed approach can solve the problems of convergency to local optima and computational burden caused by a large number of input patterns when a neural network is used for image classification. Materials and methods : Application of Hopfield neural network has been solving various optimization problems. However, major problem of mapping an image classification problem into a neural network is that network is opt to converge to local optima and its convergency toward the global solution with a standard stochastic relaxation spends much time. Therefore, to avoid local solutions and to achieve fast convergency toward a global optimization, we adopt MFA to a Hopfield network during the classification. MFA replaces the stochastic nature of simulated annealing method with a set of deterministic update rules that act on the average value of the variable. By minimizing averages, it is possible to converge to an equilibrium state considerably faster than standard simulated annealing method. Moreover, the proposed agglomerative clustering algorithm which determines the underlying clusters of the image provides initial input values of Hopfield neural network. Results : The proposed approach which uses agglomerative clustering and deterministic relaxation approach resolves the problem of local optimization and achieves fast convergency toward a global optimization when a neural network is used for MRI classification. Conclusion : In this paper, we introduce a new paradigm to classify MRI using clustering analysis and deterministic relaxation for neural network to improve the classification results.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2011.07a
• /
• pp.11-13
• /
• 2011

양방향 필터 (Bilateral filter)는 에지 보전 평활화 필터로써 디노이징, 반사 제거, 스테레오 매칭 등 다양한 분야에서 사용되고 있다. 이는 기존의 가우시안 필터에 사용되는 공간 도메인 커널 (spatial kernel)이외에 강도 도메인 커널 (range kernel)을 추가로 사용하여 비슷한 강도의 픽셀에 높은 가중치를 부여함으로써 에지를 보전하면서 평활화를 한다. 또한 양방향 필터는 비등방성 확산 필터 (Anisotropic diffusion filter)와 달리 항상 수렴을 보장한다. 따라서 본 논문에서는 고정점 반복 이론을 적용하여 양방향 필터의 수렴을 수학적으로 증명한다.

• Proceedings of the Korean Society for Noise and Vibration Engineering Conference
• /
• 1995.10a
• /
• pp.280-283
• /
• 1995

계층형 신경회로망은 학습능력이나 비선형사상능력을 가지고 있고, 그 특징을 이용하여 패턴인식이나 동정 및 제어 등에의 적용이 시도되어 성과를 올리고 있다. 현재, 그 학습법으로 널리 이용되고 있는 것이 역전파학습법으로 최급 강하법이나 공액경사법 등의 최적화 방법이 적용되고 있지만, 학습에 많은 시간이 걸리는 점, 국소적 최적해(local minima)에 해의 수렴이 이루어져 오차가 충분히 작게 되지 않는 점 등이 문제점으로 지적되고 있다. 본 논문에서는 Hu에 의해 고안된 random 탐색법과 조합된 random tabu 탐색법으로 최적결합계수를 구하는 학습알고리즘으로, 국소적 최적해에 수렴하는 것을 방지하고, 수렴정도를 개선하는 새로운 방법을 이용하여 회전기계의 이상진동진단에 적용가능성을 검토하고 오차역전파법에 의한 진단결과와 비교검토한다.

• Journal of Korea Society of Industrial Information Systems
• /
• v.2 no.2
• /
• pp.31-39
• /
• 1997

퍼지로직 컨트롤러는 파라메터의 측정이 불가능하고 수학적으로 정확하게 모델링 되지 않는 제어에 효과적으로 사용되어진다. 본 논문에서는 입력값에 대해 룩업테이블을 사용하지 않고 멤버쉽 함수를 계산하여 수렴점에서의 진동을 줄이는 방법을 제안하였다. 그리고 역추균형 제어 시스템에 적용시켜 그 타당성을 확인하였다.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2012.07a
• /
• pp.121-123
• /
• 2012

본 논문에서는 2D-to-3D 변환을 위한 기하구조 기반 깊이 추정에 적합한 소실점 검출 기법을 제안한다. 3D 공간에서 평행한 직선들은 2D 공간으로의 투시영상에서 시점에서 멀어질수록 간격이 좁아지고, 결국에는 한 점으로 수렴하게 된다. 수렴된 점을 소실점(vanishing point)이라 하고, 소실점을 거쳐 지나는 직선들을 소실선(vanishing lines)이라고 한다. 일반적으로, 인간은 소실선과 소실점을 추정한 2D 영상에서 소실점이 관찰자 시점으로부터 제일 먼 지점이라는 인식을 이용하여 깊이 정보를 인지할 수 있다. 전경영역과 배경영역 간의 경계에서는 수직성분을 가진 선들이 생성되어 올바른 소실점을 검출하는데 방해가 될 수 있다. 그렇기 때문에 본 논문에서는 수직성분을 가진 선들을 제거하여 소실점을 탐색하는 기법을 제안한다.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.8 no.5
• /
• pp.98-106
• /
• 1998

This paper proposes an efficient learning algorithm for improving the training performance of the neural network. The proposed method improves the training performance by applying the backpropagation algorithm of a global optimization method which is a hybrid of a stochastic approximation and a conjugate gradient method. The approximate initial point for f a ~gtl obal optimization is estimated first by applying the stochastic approximation, and then the conjugate gradient method, which is the fast gradient descent method, is applied for a high speed optimization. The proposed method has been applied to the parity checking and the pattern classification, and the simulation results show that the performance of the proposed method is superior to those of the conventional backpropagation and the backpropagation algorithm which is a hyhrid of the stochastic approximation and steepest descent method.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.21 no.9
• /
• pp.2349-2354
• /
• 1996

Convergence analysis on Godard's quartic (GQ) algorithm used forblind equalization is accomplished in this paper. The first main result is an explanation of the lacal behavior of the GQ algorithm around the global minimum point of the average performance functio, from which we can determine the adaptation gain. It is show that the normalized adaptation gain of the GQ algorithm should be smaller than that of the decision directed (DD) algorithm. In addition, it is observed that the GQ algorithm converges faster than the DD equalization algorithm.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.22 no.9
• /
• pp.2005-2013
• /
• 1997

The MCMA adaptive blind equalization algorithm has a excellent phase correction capabilities in addition to channel amplitude equalization, but has an inevitable error by mismatching between the original constellation points in arriving at the perfect equalization since unique new type constellation points are used as desired response instead of original constellation points and follows the slow convergence speed of CMA. In this paper, We propose an adaptive blind equalization algorithm with dual-mode, which has decision regions. Inside the decision regions, it operates as considering the moudlus of original data symbol point and outside the decision region, it operates as considerin gthe modulus of new constellation points. The proposed algorithm has a lower error in the steady state and rapid convergence speed toward steady state using the original data symbol points instead of new constellation points in the decision regions. From computer simulation, we confirm that the propposed algorithm has the performance superiority in residual ISI, convergence speed compared with the cnventional adaptive blind equalization algorithms, CMA, MCMA, Stop-and-Go algorithm.

• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.12 no.9
• /
• pp.4147-4152
• /
• 2011

Compared to the constant modulus algorithm (CMA), the modified CMA (MCMA) is easy not only to improve the steady-state performance but also to be expanded to higher-order constellations by using fewer moduli with evenly spaced. In this paper, it is shown that the MCMA is sufficient to achieve satisfactory steady-state performance by applying a variable step-size to the MCMA without switching to an hard decision-directed algorithm. Two new methods varying the step-size are proposed, and the performance improvement of the MCMA with the new methods of variable step-size is presented as compared to the CMA and the fixed step-size MCMA through computer simulations.