• Journal of the Computational Structural Engineering Institute of Korea
• /
• v.27 no.2
• /
• pp.63-70
• /
• 2014

Design sensitivity analysis and topology design optimization for a piezoelectric crystal resonator are developed. The piezoelectric crystal resonator is deformed mechanically when subjected to electric charge on the electrodes, or vice versa. The Mindlin plate theory with higher-order interpolations along thickness direction is employed for analyzing the thickness-shear vibrations of the crystal resonator. Thin electrode plates are masked on the top and bottom layers of the crystal plate in order to enforce to vibrate it or detect electric signals. Although the electrode is very thin, its weight and shape could change the performance of the resonators. Thus, the design variables are the bulk material densities corresponding to the mass of masking electrode plates. An optimization problem is formulated to find the optimal topology of electrodes, maximizing the thickness-shear contribution of strain energy at the desired motion and restricting the allowable volume and area of masking plates. The necessary design gradients for the thickness-shear frequency(eigenvalue) and the corresponding mode shape(eigenvector) are computed very efficiently and accurately using the analytical design sensitivity analysis method using the eigenvector expansion concept. Through some demonstrative numerical examples, the design sensitivity analysis method is verified to be very efficient and accurate by comparing with the finite difference method. It is also observed that the optimal electrode design yields an improved mode shape and thickness-shear energy.

• Proceedings of the Acoustical Society of Korea Conference
• /
• 1994.06c
• /
• pp.157-160
• /
• 1994

본 논문에서는 한국어 숫자음 인식에 대한 이산분포 HMM과 연속분포 HMM의 인식 성능을 비교하였다. 일반적으로 연속분포 HMM은 많은 계산량이 필요하고, 학습시 초기값이 매우 민감하다는 단점이 있지만, 이산분포 HMM의 VQ로 인한 왜곡을 제거함으로써 인식률을 향상시킬 수 있다. 여기서는 성능비교를 위해서 mel-cepstrum의 분석차수, 이산분포 HMM의 codebook 크기, 연속분포 HMM의 miture 개수등에 따른 인식성능을 비교하였다. 실험 결과 이산분포 HMM에서는 mel-cepstrum 벡터가 14차이고, codebook 크기가 64일 때 가장 좋은 성능을 나타냈으며, 연속부포 HMM에서는 mel-cepstrum 벡터가 16차이고 miture가 3개일 때 가장 좋은 결과를 얻을 수 있었다. 특히 학습 데이터의 양이 적은 경우에는 연속분포 HMM이 이산분포 HMM보다 더 좋은 인식률을 나타내었다.

• Proceedings of the IEEK Conference
• /
• 2001.09a
• /
• pp.293-296
• /
• 2001

MPEG-2 압축 방법을 이용한 비트열은 가변장 부호를 이용하기 때문에 에러에 매우 민감하다. 하나의 비트에러가 발생하더라도, 다음 동기화 부호를 찾을 때까지 매크로블록 또는 슬라이스 단위의 정보 손실을 초래하기 때문에 복원 영상의 화질 열화가 심각하다. 따라서 에러 영상의 복원을 위한 에러 은닉 기술은 복호기 쪽에 매우 중요하다. 기존에 발표된 방법들은 에러에 의한 손실이 매크로블록 단위로 발생했다는 가정 하에 손상된 매크로블록주위의 상, 하, 좌, 우 네 방향의 데이터를 이용하였다. 하지만 대부분의 심각한 에러는 슬라이스 단위로 발생하기 때문에 좌, 우의 데이터는 사용할 수 없다. 본 논문에서는 이러한 슬라이스 단위의 에러를 은닉하기에 적합한 알고리즘을 제안한다. 상, 하, 오른쪽 상단, 왼쪽 상단, 오른쪽 하단, 왼쪽 하단의 6영역의 데이터를 이용하여 두 가지 대표적인 에러 은닉 방법인 boundary matching 방식과 주변 움직임 벡터 정보를 이용한 움직임 벡터 추정 방식에 적용하였다. 실험 곁과 기존의 방법에 비해 향상된 복원 화질을 얻을 수 있었다.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2002.11a
• /
• pp.649-652
• /
• 2002

PCA 방법은 입력 차원을 수학적으로 줄일 수 있는 장점 때문에 패턴인식 부분에서 널리쓰이고 있다. 얼굴인식에서의 PCA는 학습 패턴의 분산을 최대로 하는 기저 벡터들인 고유얼굴을 만들어 얼굴인식이 필요한 영상을 이 기저 벡터에 투사시켜 이때 나온 인자들과 원래 각 개인의 대표 인자값과의 거리 비교로 얼굴을 인식하는 방법이다. 그러나 조명등의 영향에 매우 민감하며 거리값으로 얼굴을 인식하기 때문에 다양한 변화에 따라 오인식률이 높아진다. 이는 인식률을 높이고자 임계값을 높게 설정하는 과정에서 발생하는 오류이며, 이를 방지하기 위해 임계치를 낮게 설정하면 오거부율이 높아진다. 이에 본 연구에서는 PCA에 입력되는 패턴들을 사전에 비교, 분류하여 PCA 연산시에 분산과 변위를 최대한으로 가질 수 있도록 하였다. 그러하여, 기존의 PCA보다 상당히 낮은 임계값으로도 오거부율의 증가를 막았으며, 낮은 임계값 설정으로 인하여 오인식률을 낮추는 결과를 얻을 수 있었다. 이는 기존의 PCA 방법을 사용하는 인식시스템에서 종종 발생하는 허가되지 않아야 하는 외부인을 인증시키는 사례를 줄일 수 있다.

• The Journal of the Korea Contents Association
• /
• v.6 no.2
• /
• pp.65-74
• /
• 2006

The MPEG-2 video compressed bitstream is very sensitive to transmission errors due to the complex coding structure of the MPEG-2 video coding standard. If one packet is lost or received with errors, not only the current frame will be corrupted, but also errors will propagate to succeeding frames within a group of pictures. Therefore, we employ various error resilient coding/decoding techniques to protect and reduce the transmission error effects. Error concealment technique is one of them. Error concealment technique exploits spatial and temporal redundancies of the correctly received video data to conceal the corrupted video data. Motion vector recovery and compensation with the estimated motion vector is good approach to conceal the corrupted data. In this paper, we propose various error concealment algorithms based on motion vector recovery, and compare their performance to those of conventional error concealment methods.

• The Journal of the Korea Contents Association
• /
• v.7 no.2
• /
• pp.1-10
• /
• 2007

In this paper, we propose a face recognition system using AAM and Gabor feature vectors. EBGM, which is prominent among face recognition algorithms employing Gabor feature vectors, requires localization of facial feature points where Gabor feature vectors are extracted. However, localization of facial feature points employed in EBGM is based on Gator jet similarity and is sensitive to initial points. Wrong localization of facial feature points affects face recognition rate. AAM is known to be successfully applied to localization of facial feature points. In this paper, we propose a facial feature point localization method which first roughly estimate facial feature points using AAM and refine facial feature points using Gabor jet similarity-based localization method with initial points set by the facial feature points estimated from AAM, and propose a face recognition system based on the proposed localization method. It is verified through experiments that the proposed face recognition system using the combined localization performs better than the conventional face recognition system using the Gabor similarity-based localization only like EBGM.

• Economic and Environmental Geology
• /
• v.28 no.4
• /
• pp.405-414
• /
• 1995

An algorithm for monitoring geomorphological change using remote sensing data is investigated and tested using two LANDSAT TM data sets acquired over the Kyunggi Bay on April 15 1986 and September 22 1992, respectively. The algorithm exploits change vector analysis and tasseled cap transform. Although change vector analysis is effective for change detection, efficiency is decreased as the number of variables are increased. In this algorithm, we overcome the problem by utilizing the tasseled cap transform which can reduce six bands of LANDSAT TM data into only two components called Brightness and Greenness. The test results demonstrate that the algorithm is very effective in monitoring small-scaled changes over coastal area as well as significant changes in geomorphology. The resulting change vector image, however, is more sensitive to the changes occurred by human activities than by pure geological processes mainly because of relatively short time interval between two LANDSAT TM data sets.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.34 no.4C
• /
• pp.355-362
• /
• 2009

A new intra field deinterlacing algorithm with edge map in the image block is introduced. Conventional deinterlacing methods usually employ edge-based line average algorithm within pixel-by-pixel approach. However, it is sensitive to variation of intensity. To reduce this shortcoming, we proposed edge direction vector computed by edge map, and also its interpolation technique. We first introduce an edge direction vector, which is computed by Sobel mask, so that finer resolution of the edge direction can be acquired. The proposed edge direction vector oriented deinterlacer operates by identifying small pixel variations in five orientations, while weighted averaging to estimate missing pixel. According to the edge direction of the direction vector, we calculate weights on each edge direction. These weight values multiplied by the candidate deinterlaced pixels in order to successively build approximations of the deinterlaced sequence.

• Communications for Statistical Applications and Methods
• /
• v.17 no.4
• /
• pp.561-574
• /
• 2010

We studied the performance of 8 data mining algorithms including decision trees, logistic regression, LDA, QDA, Neral network, and SVM and their combinations of 2 ensemble techniques, bagging and boosting. In this study, we utilized 13 data sets with binary responses. Sensitivity, Specificity and missclassificate error were used as criteria for comparison.

• Proceedings of the KIEE Conference
• /
• 1989.07a
• /
• pp.616-621
• /
• 1989

This paper describes newly developed high performance control system for vector controlled induction motor. In order to realize this system, the authors are adopted the current controlled PWM inverter with hysteresis controller. A priori knowledge of the motor parameters is need in order to implement indirect vector control on induction motor drive systems where the position of rotor flux is estimated. This paper presents some of the relevant mathematics necessary to study parameter sensitivity in induction motor servo system. Simulation results demonstrate the predicted performance.