• 대한전자공학회:학술대회논문집
• /
• 대한전자공학회 2001년도 하계종합학술대회 논문집(3)
• /
• pp.25-28
• /
• 2001

RBFN has some problem that because the basis function isn't orthogonal to each others the number of used basis function goes to big. In this reason, the Wavelet Neural Network which uses the orthogonal basis function in the hidden node appears. In this paper, we propose the composition method of the actual function in hidden layer with the scaling function which can represent the region by which the several wavelet can be represented. In this method, we can decrease the size of the network with the pure several wavelet function. In addition to, when we determine the parameters of the scaling function we can process rough approximation and then the network becomes more stable. The other wavelets can be determined by the global solutions which is suitable for the suggested problem using the genetic algorithm and also, we use the back-propagation algorithm in the learning of the weights. In this step, we approximate the target function with fine tuning level. The complex neural network suggested In this paper is a new structure and important simultaneously in the point of handling the determination problem in the wavelet initialization.

• 전자공학회논문지CI
• /
• 제44권1호
• /
• pp.26-32
• /
• 2007

연속적 화자 분류에 있어서 분류 대상이 되는 화자에 대한 정보가 없거나 부족할 경우 정확한 연속적 분류가 어렵다. 이러한 문제를 해결하기 위해 표본 화자모델을 이용하는 방법이 제안되었는데, 이 방법을 이용하면 미리 준비된 화자의 데이터가 없이 화자모델 초기화와 화자분류가 가능해진다. 하지만 여전히 화자모델의 표본을 얻는 방법에 어려움이 따른다. 이 문제를 해결하기 위해 벡터 양자화에서 비롯된 화자 양자화를 제안한다. 유선전화 데이터를 이용한 실험에서 화자 양자화를 이용한 표본 화자모델 방법은 무작위 표본추출 방법을 이용할 경우 보다 25%의 성능 향상을 보였다.

• 대한전기학회논문지:시스템및제어부문D
• /
• 제49권7호
• /
• pp.389-398
• /
• 2000

The popular methods to check the identity of individuals include passwords and ID cards. These conventional method for user identification and authentication are not altogether reliable because they can be stolen and forgotten. As an alternative of the existing methods, biometric technology has been paid much attention for the last few decades. In this paper, we propose an efficient system for recognizing the identity of a living person by analyzing iris patterns which have a high level of stability and distinctiveness than other biometric measurements. The proposed system is based on wavelet transform and a competitive neural network with the improved mechanisms. After preprocessing the iris data acquired through a CCD camera, feature vectors are extracted by using Haar wavelet transform. LVQ(Learning Vector Quantization) is exploited to classify these feature vectors. We improve the overall performance of the proposed system by optimizing the size of feature vectors and by introducing an efficient initialization of the weight vectors and a new method for determining the winner in order to increase the recognition accuracy of LVQ. From the experiments, we confirmed that the proposed system has a great potential of being applied to real applications in an efficient and effective way.

• 한국지능시스템학회:학술대회논문집
• /
• 한국퍼지및지능시스템학회 2001년도 춘계학술대회 학술발표 논문집
• /
• pp.65-68
• /
• 2001

RBFN has some problem that because the basis function isnt orthogonal to each others the number of used basis function goes to big. In this reason, the Wavelet Neural Network which uses the orthogonal basis function in the hidden node appears. In this paper, we propose the composition method of the actual function in hidden layer with the scaling function which can represent the region by which the several wavelet can be represented. In this method, we can decrease the size of the network with the pure several wavelet function. In addition to, when we determine the parameters of the scaling function we can process rough approximation and then the network becomes more stable. The other wavelets can be determined by the global solutions which is suitable for the suggested problem using the genetic algorithm and also, we use the back-propagation algorithm in the learning of the weights. In this step, we approximate the target function with fine tuning level. The complex neural network suggested in this paper is a new structure and important simultaneously in the point of handling the determination problem in the wavelet initialization.

• 한국통신학회논문지
• /
• 제28권1B호
• /
• pp.27-32
• /
• 2003

본 논문에서는 고속 네트웍 기반의 분산 병렬 시스템에서 다양한 내재 병렬 형태를 갖는 프로그램의 효과적인 수행을 위한 유전자 알고리즘 기반의 태스크 스케쥴링 방법(Genetic Algorithm based Task Scheduling GATS)을 제안한다. 분산병렬 시스템은 고속 네트웍을 통해 연결되어진 다수의 범용, 병렬, 벡터 컴퓨터들로 구성되어진다. 분산병렬 처리의 목적은 다양한 내재 병렬 형태를 갖는 연산 집약적인 문제들을 다수의 고성능 및 병렬 컴퓨터들의 각기 다른 능력을 최대한 이용하여 해결함에 있다 분산병렬 시스템에서 스케쥴링을 통하여 더 많은 속도향상을 얻기 위해서는 시스템간의 부하 균형보다는 태스크와 병렬 컴퓨터간의 병렬특성의 일치가 주의 깊게 다루어져야 하며 태스크의 이동으로 인한 통신 오버헤드가 최소화되어야 한다 본 논문에서는 유전자 알고리즘의 동작이 병렬 특성을 감안하여 이루어질 수 있도록 초기화 방법과 지식 기반의 mutation 방법을 제안한다.

• 제어로봇시스템학회논문지
• /
• 제4권2호
• /
• pp.256-263
• /
• 1998

Among process synthesis problems, the heat exchanger network (HEN) has been subjected to the most concentrated effort because this kind of problems was well defined for solving it and 20-30% energy savings could be realized in the present chemical processes. In this paper, we use an evolutionary approach for HEN synthesis because this approach can overcome the local optimum and combine some heuristic rules. The basic evolutionary approach is composed of three parts, that is, initialization step, growth step and mutation step, as in the simulated annealing and genetic algorithm. This algorithm uses the ecological rule that a better cell will live and worse cell should decompose after repeated generations. With this basic concept, a new procedure is developed and a more efficient method is proposed to generate initial solutions. Its effectiveness is shown using test examples.

• 로봇학회논문지
• /
• 제12권1호
• /
• pp.33-41
• /
• 2017

We present a region-based approach for accurate pose estimation of small mechanical components. Our algorithm consists of two key phases: Multi-view object co-segmentation and pose estimation. In the first phase, we explain an automatic method to extract binary masks of a target object captured from multiple viewpoints. For initialization, we assume the target object is bounded by the convex volume of interest defined by a few user inputs. The co-segmented target object shares the same geometric representation in space, and has distinctive color models from those of the backgrounds. In the second phase, we retrieve a 3D model instance with correct upright orientation, and estimate a relative pose of the object observed from images. Our energy function, combining region and boundary terms for the proposed measures, maximizes the overlapping regions and boundaries between the multi-view co-segmentations and projected masks of the reference model. Based on high-quality co-segmentations consistent across all different viewpoints, our final results are accurate model indices and pose parameters of the extracted object. We demonstrate the effectiveness of the proposed method using various examples.

• ETRI Journal
• /
• 제43권1호
• /
• pp.120-132
• /
• 2021

A frequency diverse array (FDA) multiple-input multiple-output (MIMO) radar employs a small frequency increment across transmit elements to produce an angle-range-dependent beampattern for target angle and range detection. The joint angle and range estimation problem is a trilinear model. The traditional trilinear alternating least square (TALS) algorithm involves high computational load due to excessive iterations. We propose a fast-convergence trilinear decomposition (FC-TD) algorithm to jointly estimate FDA-MIMO radar target angle and range. We first use a propagator method to obtain coarse angle and range estimates in the data domain. Next, the coarse estimates are used as initialized parameters instead of the traditional TALS algorithm random initialization to reduce iterations and accelerate convergence. Finally, fine angle and range estimates are derived and automatically paired. Compared to the traditional TALS algorithm, the proposed FC-TD algorithm has lower computational complexity with no estimation performance degradation. Moreover, Cramer-Rao bounds are presented and simulation results are provided to validate the proposed FC-TD algorithm effectiveness.

• Structural Engineering and Mechanics
• /
• 제77권5호
• /
• pp.613-626
• /
• 2021

In this paper, an enhanced Violation-based Sensitivity analysis and Border-Line Adaptive Sliding Technique (ViS-BLAST) will be utilized for optimization of some well-known structural and mechanical engineering problems. ViS-BLAST has already been introduced by the authors for solving truss optimization problems. For those problems, this method showed a satisfactory enactment both in speed and efficiency. The Enriched ViS-BLAST or EVB is introduced to be vastly applicable to any solvable constrained optimization problem without any specific initialization. It uses one-directional step-wise searching technique and mostly limits exploration to the vicinity of FNF border and does not explore the entire design space. It first enters the feasible region very quickly and keeps the feasibility of solutions. For doing this important, EVB groups variables for specifying the desired searching directions in order to moving toward best solutions out or inside feasible domains. EVB was employed for solving seven numerical engineering design problems. Results show that for problems with tiny or even complex feasible regions with a larger number of highly non-linear constraints, EVB has a better performance compared to some records in the literature. This dominance was evaluated in terms of the feasibility of solutions, the quality of optimum objective values found and the total number of function evaluations performed.

• Current Optics and Photonics
• /
• 제5권4호
• /
• pp.362-369
• /
• 2021

Absolute position sensors are widely used in machine tools and precision measuring instruments because measurement errors are not accumulated, and position measurements can be performed without initialization. The laser speckle-based absolute position sensor, in particular, has advantages in terms of simple system configuration and high measurement accuracy. Unlike traditional absolute position sensors, it does not require an expensive physical length scale; instead, it uses a laser speckle image database to measure a moving surface position. However, there is a problem that a huge database is required to store information in all positions on the surface. Conversely, reducing the size of the database also decreases the accuracy of position measurements. Therefore, in this paper, we propose a new method to measure the surface position with high precision while reducing the size of the database. We use image stitching and approximation methods to reduce database size and speed up measurements. The absolute position error of the proposed method was about 0.27 ± 0.18 ㎛, and the average measurement time was 25 ms.