• Journal of KIISE:Software and Applications
• /
• v.30 no.7_8
• /
• pp.713-718
• /
• 2003

The CONDENSATION (Conditional Density Propagation) algorithm has a robust tracking performance and suitability for real-time implementation. However, the CONDENSATION tracker has some difficulties with real-time implementation for multiple people tracking since it requires very complicated shape modeling and a large number of samples for precise tracking performance. Further, it shows a poor tracking performance in the case of close or partially occluded people. To overcome these difficulties, we present three improvements: First, we construct effective templates of people´s shapes using the SOM (Self-Organizing Map). Second, we take the discrete HMM (Hidden Markov Modeling) for an accurate dynamical model of the people´s shape transition. Third, we use the competition rule to separate close or partially occluded people effectively. Simulation results shows that the proposed CONDENSATION algorithm can achieve robust and real-time tracking in the image sequences of a crowd of people.

• The KIPS Transactions:PartB
• /
• v.10B no.1
• /
• pp.33-40
• /
• 2003

Data mining plays an important role in a knowledge discovery process and various algorithms of data mining can be selected for the specific purpose. Most of traditional hierachical clustering methode are suitable for processing small data sets, so they difficulties in handling large data sets because of limited resources and insufficient efficiency. In this study we propose a hybrid neural networks clustering technique, called PPC for Pre-Post Clustering that can be applied to large data sets and find unknown patterns. PPC combinds an artificial intelligence method, SOM and a statistical method, hierarchical clustering technique, and clusters data through two processes. In pre-clustering process, PPC digests large data sets using SOM. Then in post-clustering, PPC measures Similarity values according to cohesive distances which show inner features, and adjacent distances which show external distances between clusters. At last PPC clusters large data sets using the simularity values. Experiment with UCI repository data showed that PPC had better cohensive values than the other clustering techniques.

• Proceedings of the Korean Information Science Society Conference
• /
• 2012.06a
• /
• pp.325-327
• /
• 2012

스마트 홈은 단순한 가정 내 네트워크 연결이 아닌 주택(건물)내의 정보 기술 요소를 구현하는 토털 홈 정보 제어 시스템 서비스, 솔루션을 총칭한다. 현재는 언제, 어디서, 어떤 기기로건 인터넷에 접속할 수 있는 유비쿼터스(Ubiquitous) 시대이자, 개별 사물들이 인터넷에 연결되어 스스로 필요한 정보를 주고받게 될 시대가 도래함에 따라 사람들의 주요 생활공간에서도 활용도가 점차 커지는 것이다. 수시로 변화하는 상황에 적응하며 정확도가 높은 스마트 서비스의 제공을 위해서는 사용자의 의도에 부합하는 Semantic-Context 정보생성을 위한 SOM(Self-Organizing Map)추론 방식의 알고리즘과 정보의 의미화로 다양한 서비스를 지원할 수 있는 인프라 대비 최대 서비스가 요구된다. 이에 따라 본 논문에서는 스마트 홈에서 이종 가전기기들의 상황정보를 센서 데이터로부터 추출하여 사용자 맞춤형 서비스를 제공하기 위한 SOM 추론 기반의 스마트 홈을 설계한다.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.32 no.9C
• /
• pp.861-866
• /
• 2007

A clustering algorithm for Gaussian Probability Distribution Function (GPDF) data called Centroid Neural Network with a Bhattacharyya Kernel (BK-CNN) is proposed in this paper. The proposed BK-CNN is based on the unsupervised competitive Centroid Neural Network (CNN) and employs a kernel method for data projection. The kernel method adopted in the proposed BK-CNN is used to project data from the low dimensional input feature space into higher dimensional feature space so as the nonlinear problems associated with input space can be solved linearly in the feature space. In order to cluster the GPDF data, the Bhattacharyya kernel is used to measure the distance between two probability distributions for data projection. With the incorporation of the kernel method, the proposed BK-CNN is capable of dealing with nonlinear separation boundaries and can successfully allocate more code vector in the region that GPDF data are densely distributed. When applied to GPDF data in an image classification probleml, the experiment results show that the proposed BK-CNN algorithm gives 1.7%-4.3% improvements in average classification accuracy over other conventional algorithm such as k-means, Self-Organizing Map (SOM) and CNN algorithms with a Bhattacharyya distance, classed as Bk-Means, B-SOM, B-CNN algorithms.

• Journal of the Korean Society for information Management
• /
• v.25 no.2
• /
• pp.49-68
• /
• 2008

In many information retrieval experiments, the data fusion techniques have been used to achieve higher effectiveness in comparison to the single evidence-based retrieval. However, there had not been many image retrieval studies using the data fusion techniques especially in combining retrieval results based on multiple retrieval methods. In this paper, we describe how the image retrieval effectiveness can be improved by combining two sets of the retrieval results using the Sobel operator-based edge detection and the Self Organizing Map(SOM) algorithms. We used the clip art images from a commercial collection to develop a test data set. The main advantage of using this type of the data set was the clear cut relevance judgment, which did not require any human intervention.

• 기계와재료
• /
• v.21 no.3
• /
• pp.62-73
• /
• 2009

전력기기에서 발생하는 부분방전을 정확히 측정하고 이를 올바르게 해석하는 작업은 신뢰성 있는 진단법을 개발하고 이를 현장에 적용하는데 있어 대단히 중요하다. 측정된 고주파 데이터를 패턴 분석이 가능한 형태로 가공하는 전처리 과정을 수행하고, 가공된 데이터를 패턴인식을 통하여 기존의 각 노이즈 및 부분방전 패턴과 비교하여 실제 측정된 데이터가 어떤 부분방전 패턴인지 판단한다. 패턴 인식 처리 방법으로는 컴퓨터 분야 신경회로망의 BP 알고리즘과 SOM 알고리즘이 널리 사용되고 있으며 본 연구에서는 TF-MAP, PRPDA, EBP 알고리즘을 이용하여 부분방전 패턴인식 기술 개발에 관한 연구를 수행하였다.

• Journal of the Korea Society of Computer and Information
• /
• v.10 no.5 s.37
• /
• pp.77-86
• /
• 2005

In this paper, we designed a learning algorithm of LVQ that extracts classification errors and learns ones and improves classification performance. The proposed LVQ learning algorithm is the learning Networks which is use SOM to learn initial reference vectors and out-star learning algorithm to determine the class of the output neurons of LVQ. To extract pattern vectors which cause classification errors, we proposed the error-cause condition, which uses that condition and constructed the pattern vector space which consists of the input pattern vectors that cause the classification errors and learned these pattern vectors , and improved performance of the pattern classification. To prove the performance of the proposed learning algorithm, the simulation is performed by using training vectors and test vectors that are Fisher' Iris data and EMG data, and classification performance of the proposed learning method is compared with ones of the conventional LVQ, and it was a confirmation that the proposed learning method is more successful classification than the conventional classification.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.31 no.2C
• /
• pp.208-218
• /
• 2006

In this paper, we introduce the creation methods of attack detection model using data mining technologies that can classify the latest attack types, and can detect the modification of existing attacks as well as the novel attacks. Also, we evaluate comparatively these attack detection models in the view of detection accuracy and detection time. As the important factors for creating detection models, there are data, attribute, and detection algorithm. Thus, we used NetFlow data gathered at the real network, and KDD Cup 1999 data for the experiment in large quantities. And for attribute selection, we used a heuristic method and a theoretical method using decision tree algorithm. We evaluate comparatively detection models using a single supervised/unsupervised data mining approach and a combined supervised data mining approach. As a result, although a combined supervised data mining approach required more modeling time, it had better detection rate. All models using data mining techniques could detect the attacks within 1 second, thus these approaches could prove the real-time detection. Also, our experimental results for anomaly detection showed that our approaches provided the detection possibility for novel attack, and especially SOM model provided the additional information about existing attack that is similar to novel attack.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.11 no.3
• /
• pp.223-230
• /
• 2001

Since self-organizing map (SOM) preserves the topology of ordering in input spaces and trains itself by unsupervised algorithm, it is Llsed in many areas. However, SOM has a shortcoming: structure cannot be easily detcrmined without many trials-and-errors. Structure-adaptive self-orgnizing map (SASOM) which can adapt its structure as well as its weights overcome the shortcoming of self-organizing map: SASOM makes use of structure adaptation capability to place the nodes of prototype vectors into the pattern space accurately so as to make the decision boundmies as close to the class boundaries as possible. In this scheme, the initialization of weights of newly adapted nodes is important. This paper proposes a method which optimizes SASOM with genetic algorithm (GA) to determines the weight vector of newly split node. The leanling algorithm is a hybrid of unsupervised learning method and supervised learning method using LVQ algorithm. This proposed method not only shows higher performance than SASOM in terms of recognition rate and variation, but also preserves the topological order of input patterns well. Experiments with 2D pattern space data and handwritten digit database show that the proposed method is promising.

• Proceedings of the Korean Institute of Intelligent Systems Conference
• /
• 2006.11a
• /
• pp.179-182
• /
• 2006

현재 감정을 인식할 수 있는 방법으로는 음성, 뇌파, 심박, 표정 등 많은 방법들이 존재한다. 본 논문은 이러한 방법 중 음성 신호를 이용한 방법으로써 특징들은 크게 피치, 에너지, 포만트 3가지 특징 점을 고려하였으며 이렇게 다양한 특징들을 사용하는 이유는 아직 획기적인 특징점이 정립되지 않았기 때문이며 이러한 선택의 문제를 해결하기 위해 본 논문에서는 특징 선택 방법 중 Multi Feature Selection(MFS) 방법을 사용하였으며 학습 알고리즘은 Self Organizing Map 알고리즘을 이용하여 음성 신호의 감정 특징 패턴을 분류하는 방법을 제안한다.