• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제9권2호
• /
• pp.742-762
• /
• 2015

In this study, we propose a distance metric learning approach called discriminant metric learning (DML) for face verification, which addresses a binary-class problem for classifying whether or not two input images are of the same subject. The critical issue for solving this problem is determining the method to be used for measuring the distance between two images. Among various methods, the large margin nearest neighbor (LMNN) method is a state-of-the-art algorithm. However, to compensate the LMNN's entangled data distribution due to high levels of appearance variations in unconstrained environments, DML's goal is to penalize violations of the negative pair distance relationship, i.e., the images with different labels, while being integrated with LMNN to model the distance relation between positive pairs, i.e., the images with the same label. The likelihoods of the input images, estimated using DML and LMNN metrics, are then weighted and combined for further analysis. Additionally, rather than using the k-nearest neighbor (k-NN) classification mechanism, we propose a verification mechanism that measures the correlation of the class label distribution of neighbors to reduce the false negative rate of positive pairs. From the experimental results, we see that DML can modify the relation of negative pairs in the original LMNN space and compensate for LMNN's performance on faces with large variances, such as pose and expression.

• Communications for Statistical Applications and Methods
• /
• 제15권6호
• /
• pp.977-991
• /
• 2008

본 논문에서는 연속형 변수에 대한 결합확률분포를 추정하지 않고도 상호정보(MI) 추정량을 구할 수 있는 k-최근접이웃 기반방법에 대하여 연구하였다. 변수가 동일한 값들을 가지는 경우 k-최근접이웃을 구할 때 생기는 문제점을 해결하기 위하여 지터링(jittering)과 붓스트랩(bootstrap) 방법을 제안하였다. 몬테칼로 모의실험과 실제 데이터에 대한 실험을 수행한 결과, k=1과 같이 작은 값을 사용한 k-최근접이웃 기반방법에 의해 효율적인 MI 추정량을 구할 수 있었다. k-최근접이웃 기반방법은 연속형 설명변수, 범주형 또는 연속형인 목적변수 형태의 데이터에 적용할 수 있으며, 목적변수에 영향을 주는 중요한 설명변수의 순서를 구할 수 있을 뿐만 아니라 다차원에도 적용할 수 있기 때문에 중요변수의 집합을 구하는 변수 선택(feature subset selection) 문제에도 적용할 수 있다.

• 한국군사과학기술학회지
• /
• 제10권4호
• /
• pp.88-96
• /
• 2007

This paper discusses the fast modulation classifiers capable of classifying both analog and digital modulation signals in wireless communications applications. A total of 7 statistical signal features are extracted and used to classify 9 modulated signals. In this paper, we investigate the performance of the two types of fast modulation classifiers (i.e. 2 nearest neighbor classifiers and 2 minimum distance classifiers) and compare the performance of these classifiers with that of the state of the art for the existing classification methods such as SVM Classifier. Computer simulations indicate good performance on an AWGN channel, even at low signal-to-noise ratios, in case of minimum distance classifiers (MDC for short) and k nearest neighbor classifiers (kNNC for short). Besides a good performance, these type classifiers are considered as ideal candidate to adapt real-time software radio because of their fast modulation classification capability.

• 한국인터넷방송통신학회논문지
• /
• 제17권6호
• /
• pp.11-17
• /
• 2017

본 논문에서는 멀웨어 탐지 방법으로 Opcode (operation code)와 실행 파일에서 추출한 Windows API Call로 구성된 특징 벡터를 사용하는 방법을 제안한다. 먼저 PE 파일에서 추출한 opcode와 windows API로 특징 벡터를 구성하고 Bernoulli Naïve Bayes과 K-Nearest Neighbor 분류기 알고리즘을 사용하여 성능을 각각 측정하였다. 실험결과, 제안한 방법과 KNN 분류기를 사용하여 분류하면 95.21%의 멀웨어 탐지 정확도를 얻을 수 있었다. 결과적으로 기존의 Opcode 또는 Windows API 호출 중 하나만 사용하는 방법보다 제안한 방법이 멀웨어 탐지 정확도에서 높은 성능을 보인다.

• 한국정밀공학회지
• /
• 제19권3호
• /
• pp.52-58
• /
• 2002

The purpose of condition monitoring and fault diagnosis is to detect faults occurring in machinery in order to improve the level of safety in plants and reduce operational and maintenance costs. The recognition performance is important not only to gain a high recognition rate bur a1so to minimize the diagnosis failures error rate by using off effective rejection module. We examined the problem of performance evaluation for the rejection scheme considering the accuracy of individual c1asses in order to increase the recognition performance. We use the Smith's method among the previous studies related to rejection method. Nearest neighbor classifier is used for classifying the machine conditions from the vibration signals. The experiment results for the performance evaluation of rejection show the modified optimum rejection method is superior to others.

• 한국안전학회지
• /
• 제15권2호
• /
• pp.1-7
• /
• 2000

This paper presents the approach involving fault detection and diagnosis of gears using pattern recognition and Wavelet transform. It describes result of the comparison between KDI (Kullback Discrimination Information) with the nearest neighbor classification rule as one of pattern recognition methods and Wavelet transform to know a way to detect and diagnosis of gear damages experimentally. To model the damages 1) Normal (no defect), 2) one tooth is worn out, 3) All teeth faces are worn out 4) One tooth is broken. The vibration sensor was attached on the bearing housing. This produced the total time history data that is 20 pieces of each condition. We chose the standard data and measure distance between standard and tested data. In Wavelet transform analysis method, the time series data of magnitude in specified frequency (rotary and mesh frequency) were earned. As a result, the monitoring system using Wavelet transform method and KDI with nearest neighbor classification rule successfully detected and classified the damages from the experimental data.

• 대한안전경영과학회지
• /
• 제11권1호
• /
• pp.27-33
• /
• 2009

A study of fracture to material is getting interest in nuclear and aerospace industry as a viewpoint of safety. Acoustic emission (AE) is a non-destructive testing and new technology to evaluate safety on structures. In previous research continuously, all tensile tests on the pre-defected coupons were performed using the universal testing machine, which machine crosshead was move at a constant speed of 5mm/min. This study is to evaluate an AE source characterization of SM45C steel by using k-nearest neighbor classifier, k-NNC. For this, we used K-means clustering as an unsupervised learning method for obtained multi -variate AE main data sets, and we applied k-NNC as a supervised learning pattern recognition algorithm for obtained multi-variate AE working data sets. As a result, the criteria of Wilk's $\lambda$, D&B(Rij) & Tou are discussed.

• 전자공학회논문지B
• /
• 제28B권11호
• /
• pp.852-858
• /
• 1991

An adaptive hierarchical classifier(AHCL) for Korean character recognition using a neural net is designed. This classifier has two neural nets: USACL (Unsupervised Adaptive Classifier) and SACL (Supervised Adaptive Classifier). USACL has the input layer and the output layer. The input layer and the output layer are fully connected. The nodes in the output layer are generated by the unsupervised and nearest neighbor learning rule during learning. SACL has the input layer, the hidden layer and the output layer. The input layer and the hidden layer arefully connected, and the hidden layer and the output layer are partially connected. The nodes in the SACL are generated by the supervised and nearest neighbor learning rule during learning. USACL has pre-attentive effect, which perform partial search instead of full search during SACL classification to enhance processing speed. The input of USACL and SACL is a directional edge feature with a directional receptive field. In order to test the performance of the AHCL, various multi-font printed Hangul characters are used in learning and testing, and its processing its speed and and classification rate are compared with the conventional LVQ(Learning Vector Quantizer) which has the nearest neighbor learning rule.

• 토지주택연구
• /
• 제1권1호
• /
• pp.1-7
• /
• 2010

Case-based reasoning (CBR) method can make estimators understand the estimation process more clearly. Thus, CBR is widely used as a methodology for cost estimation. In CBR, the quality of case retrieval affects the relevance of retrieved cases and hence the overall quality of the reminding capability of CBR system. Thus, it is essential to retrieve relevant past cases for establishing a robust CBR system. Case retrieval needs the following tasks to obtain appropriate case(s); indexing, search, and matching (Aamodt and Plaza 1994). However, the previous CBR researches mostly deal with matching process that has limits such as accuracy and efficiency of case retrieval. In order to address this issue, this research presents a CBR cost model for building projects that has two-step retrieval process: decision tree and nearest neighbor methods. Specifically, the proposed cost model has indexing, search and matching modules. Features in the model are divided into shape-based and scale-based attributes. Based on these, decision tree is established for facilitating the search task and nearest neighbor method was utilized for matching task. In regard to applying nearest neighbor method, attribute weights are assigned using GA optimization and similarity is calculated using the principle of distance measuring. Thereafter, the proposed CBR cost model is developed using 174 cases and validated using 12 test cases.

• 한국방송∙미디어공학회:학술대회논문집
• /
• 한국방송공학회 2015년도 하계학술대회
• /
• pp.122-124
• /
• 2015

본 논문에서는 Microsoft 에서 나온 여러 대의 Kinect 를 이용하여 Free Viewpoint TV System 을 구현해 보고자 한다. Kinect 로부터 얻어진 색상 영상과 깊이 영상을 통하여, 실시간으로 두 대의 카메라 사이에서의 가상시점에서 영상이 출력되는 시스템을 설계한다. 또한, 여러 대의 Kinect 를 이용할 때, 간섭현상으로 인해 IR 패턴을 제대로 인식하지 못하여 홀이 생성되는 문제점을 확인하고, Nearest Neighbor 방식과 Inpainting 기법을 사용하여 홀을 제거하는 방식을 소개한다. 실험 결과, 홀의 주변과 비슷한 값으로 홀을 채울 수 있었지만, 홀의 크기에 따라 Edge 경계가 부정확해 지는 현상을 확인할 수 있다.