• Annual Conference of KIPS
• /
• 2000.04a
• /
• pp.1051-1055
• /
• 2000

본 연구에서는 내용기반 영상 데이터 검색을 위하여 변환 영역에서 위치 정보와 주파수 정보를 가지는 웨이블릿 성질을 이용하여 객체들의 특징을 추출하는 방안인 Vector Quantization 을 이용한 영상을 검색하는 방안을 제시한다. 내용기반 영상 검색의 주요 특징들은 색상, 질감, 그리고 영상의 공간적인 특징을 고려한 특징 값 등이 사용된다. 이러한 영상의 특징들을 어떻게 결합하고 특징 추출을 하느냐에 따라 검색의 효율성에 영향을 준다. 따라서 본 연구에서는 영상의 위치 정보와 주파수 정보를 가지는 웨이블릿 변환 후 얻어지는 저대역 부밴드에서의 공간적인 특성을 고려한 특징 값을 이용하여 Vector Quantization 알고리즘에 의해 정지영상의 객체 대표 특징들을 빠르게 검색하고자 한다. 본 연구에서는 Haar Wavelet 과 Vector Quantization 에서 색상과 질감의 가중치를 적용하고자 한다.

• International Journal of Computer Science & Network Security
• /
• v.21 no.6
• /
• pp.1-6
• /
• 2021

Recently, pre-trained convolutional neural network CNNs have been widely used and applied for medical image classification. These models can utilised in three different ways, for feature extraction, to use the architecture of the pre-trained model and to train some layers while freezing others. In this study, the ResNet18 pre-trained CNNs model is used for feature extraction, followed by the support vector machine for multiple classes to classify medical images from multi-classes, which is used as the main classifier. Our proposed classification method was implemented on Kvasir and PH2 medical image datasets. The overall accuracy was 93.38% and 91.67% for Kvasir and PH2 datasets, respectively. The classification results and performance of our proposed method outperformed some of the related similar methods in this area of study.

• Annual Conference on Human and Language Technology
• /
• 2020.10a
• /
• pp.427-432
• /
• 2020

딥러닝(Deep-learning) 기반의 자연어 이해(Natural Language Understanding) 기술들은 최근에 상당한 성과를 성취했다. 하지만 딥러닝 기반의 자연어 이해 기술들은 내적인 동작들과 결정에 대한 근거를 설명하기 어렵다. 본 논문에서는 벡터를 그래프로 변환함으로써 신경망의 내적인 의미 표현들을 설명할 수 있도록 한다. 먼저 인간과 기계 모두가 이해 가능한 표현방법의 하나로 그래프를 주요 표현방법으로 선택하였다. 또한 그래프의 구성요소인 노드(Node) 및 엣지(Edge)의 결정을 위한 Element-Importance Inverse-Semantic-Importance(EI-ISI) 점수와 Element-Element-Correlation(EEC) 점수를 심층신경망의 훈련방법 중 하나인 드랍아웃(Dropout)을 통해 계산하는 방법을 제안한다. 다양한 실험들을 통해, 본 연구에서 제안한 벡터-그래프(Vector2graph) 변환 프레임워크가 성공적으로 벡터의 의미정보를 유지하면서도, 설명 가능한 그래프를 생성함을 보인다. 더불어, 그래프 기반의 새로운 시각화 방법을 소개한다.

• Journal of the Korea Computer Graphics Society
• /
• v.29 no.3
• /
• pp.159-166
• /
• 2023

This paper investigates effective methods for implementing motion matching, which is actively used in real-time motion generation applications. The success of motion matching heavily hinges on its simple definition of a feature vector, yet this very definition can introduce significant variance in the outcomes. Our research focuses on identifying the optimal combination of feature vectors that effectively generates desired trajectories in locomotion generation. To this end, we experimented with a range of feature vector combinations and performed an in-depth error analysis to evaluate the results.

• Journal of the Korean Institute of Telematics and Electronics
• /
• v.25 no.5
• /
• pp.550-558
• /
• 1988

This paper presents laplacian pyramid vector quantization (LPVQ) approach in which a vector quantizer is used to encode a series of quasi-bandpassed images generated by the laplacian pyramid processing. Performance of the LPVQ is compared to those of DCT domain methods at the same bit rate via computer simulations. Experimental results show that the PSNR's (peak signal-to-noise ratio) for the LPVQ are almost the same as those of the DCT based methods. However, subjective study indicates the LPVQ obtains slightly higher scores than the DCT based techniques.

• Proceedings of the KIPE Conference
• /
• 1999.07a
• /
• pp.390-393
• /
• 1999

For high performance ac drives, the speed sensorless vector control and a speed control algorithm base on the Fuzzy PI controller have received increasing attention. A Fuzzy PI controller is used for robust and fast speed control and space vector modulation method is used for PWM wave generation in this proposed system. The computer simulation results show that the proposed controller are more excellent control characteristics than conventional PI controller in transient-state response.

• Communications for Statistical Applications and Methods
• /
• v.19 no.5
• /
• pp.655-662
• /
• 2012

Classification is an important research field in pattern recognition with high-dimensional predictors. The support vector machine(SVM) is a penalized feature selector and classifier. It is based on the hinge loss function, the non-convex penalty function, and the smoothly clipped absolute deviation(SCAD) suggested by Fan and Li (2001). We developed the algorithm for the multiclass SVM with the SCAD penalty function using the local quadratic approximation. For multiclass problems we compared the performance of the SVM with the $L_1$, $L_2$ penalty functions and the developed method.

• Bulletin of the Korean Mathematical Society
• /
• v.42 no.2
• /
• pp.279-294
• /
• 2005

We study a real hypersurface M satisfying $L_{\xi}S=0\;and\;R_{\xi}S=SR_{\xi}$ in a complex hyperbolic space $H_n\mathbb{C}$, where S is the Ricci tensor of type (1,1) on M, $L_{\xi}\;and\;R_{\xi}$ denotes the operator of the Lie derivative and the Jacobi operator with respect to the structure vector field e respectively.

• Proceedings of the Korean Institute of Intelligent Systems Conference
• /
• 2007.11a
• /
• pp.217-220
• /
• 2007

In this paper, we present a new anchor shot detection system which is a core step of the preprocessing process for the news video analysis. The proposed system is composed of four modules and operates sequentially: 1) skin color detection module for reducing the candidate face regions; 2) face detection module for finding the key-frames with a facial data; 3) vector representation module for the key-frame images using a non-negative matrix factorization; 4) anchor shot detection module using a support vector data description. According to our computer experiments, the proposed system shows not only the comparable accuracy to the recent other results, but also more faster detection rate than others.

• 제어로봇시스템학회:학술대회논문집
• /
• 2005.06a
• /
• pp.2542-2547
• /
• 2005

In this paper, we propose a support vector machine (SVM) hyper and kernel parameter optimization method which is based on minimizing radius/margin bound which is a kind of estimation of leave-one-error. This method uses dynamic encoding algorithm for search (DEAS) and gradient information for better optimization performance. DEAS is a recently proposed optimization algorithm which is based on variable length binary encoding method. This method has less computation time than genetic algorithm (GA) based and grid search based methods and better performance on finding global optimal value than gradient based methods. It is very efficient in practical applications. Hand-written letter data of MNI steel are used to evaluate the performance.