• Proceedings of the Korean Institute of Intelligent Systems Conference
• /
• 2006.11a
• /
• pp.61-64
• /
• 2006

SVDD(support vector data description)는 one-class 서포트 벡터 학습 방법론 중 하나로 비정상 물체에서 정상 데이터를 구분하기 위해서 특징 공간(feature space)에서 정의된 구를 이용하는 전략을 쓰는 방법론이다. 하지만 SVDD는 모든 데이터에 대해서 같은 중요도를 부가하는 단점을 가지고 있다. 최근에, 이런 문제점을 보완하기 위해 데이터의 밀도 분포에 따라서 중요도를 다르게 부가하는 D-SVDD(density-induced support vector data description) 방법론이 발표되었고, 아직도 많은 연구가 진행되고 있다. 본 논문에서는 D-SVDD를 이용해서 노이즈가 섞인 비정상 데이터를 노이즈가 제거된 정상 데이터로 복원하는 방법에 대해서 논한다. 특히, 본 논문에서 제안하는 방법론을 다른 방법론과 비교하여 본 논문의 방법론의 효용성에 대해서 다룬다.

• International Journal of Fuzzy Logic and Intelligent Systems
• /
• v.7 no.1
• /
• pp.1-6
• /
• 2007

The SVDD (support vector data description) is one of the most well-known one-class support vector learning methods, in which one tries the strategy of utilizing balls defined on the feature space in order to distinguish a set of normal data from all other possible abnormal objects. Recently, with the objective of generalizing the SVDD which treats all training data with equal importance, the so-called D-SVDD (density-induced support vector data description) was proposed incorporating the idea that the data in a higher density region are more significant than those in a lower density region. In this paper, we consider the problem of further improving the D-SVDD toward the use of a partial reference set for testing, and propose an LMI (linear matrix inequality)-based optimization approach to solve the improved version of the D-SVDD problems. Our approach utilizes a new class of density-induced distance measures based on the RSDE (reduced set density estimator) along with the LMI-based mathematical formulation in the form of the SDP (semi-definite programming) problems, which can be efficiently solved by interior point methods. The validity of the proposed approach is illustrated via numerical experiments using real data sets.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.24 no.5
• /
• pp.518-528
• /
• 2014

In this paper, we propose a method to recognize the action direction of human by developing 4D space-time (4D-ST, [x,y,z,t]) features. For this, we propose 4D space-time interest points (4D-STIPs, [x,y,z,t]) which are extracted using 3D space (3D-S, [x,y,z]) volumes reconstructed from images of a finite number of different views. Since the proposed features are constructed using volumetric information, the features for arbitrary 2D space (2D-S, [x,y]) viewpoint can be generated by projecting the 3D-S volumes and 4D-STIPs on corresponding image planes in training step. We can recognize the directions of actors in the test video since our training sets, which are projections of 3D-S volumes and 4D-STIPs to various image planes, contain the direction information. The process for recognizing action direction is divided into two steps, firstly we recognize the class of actions and then recognize the action direction using direction information. For the action and direction of action recognition, with the projected 3D-S volumes and 4D-STIPs we construct motion history images (MHIs) and non-motion history images (NMHIs) which encode the moving and non-moving parts of an action respectively. For the action recognition, features are trained by support vector data description (SVDD) according to the action class and recognized by support vector domain density description (SVDDD). For the action direction recognition after recognizing actions, each actions are trained using SVDD according to the direction class and then recognized by SVDDD. In experiments, we train the models using 3D-S volumes from INRIA Xmas Motion Acquisition Sequences (IXMAS) dataset and recognize action direction by constructing a new SNU dataset made for evaluating the action direction recognition.