• Journal of Convergence for Information Technology
• /
• v.10 no.9
• /
• pp.15-21
• /
• 2020

This paper presents the outlier detection algorithm in the estimation method of a source location and velocity based on two-step weighted least-squares method using time difference of arrival(TDOA) and frequency difference of arrival(FDOA) data. Since the accuracy of the estimated location and velocity of a moving source can be reduced by the outliers of TDOA and FDOA data, it is important to detect and remove the outliers. In this paper, the method to find the minimum inlier data and the method to determine whether TDOA and FDOA data are included in inliers or outliers are presented. The results of numerical simulations show that the accuracy of the estimated location and velocity is improved by removing the outliers of TDOA and FDOA data.

• The KIPS Transactions:PartB
• /
• v.15B no.2
• /
• pp.103-112
• /
• 2008

Image-based motion capture technology is often used in making realistic computer animation. In this paper we try to implement image-based motion rendering by fixing a camera to a PC. Existing image-based rendering algorithms have disadvantages of high computational burden or low accuracy. The former disadvantage causes too long making-time of an animation. The latter disadvantage degrades reality in making realistic animation. To compensate for those disadvantages of the existing approaches, this paper presents an image-based motion rendering algorithm with low computational load and high estimation accuracy. In the proposed approach, an incremental motion rendering algorithm with low computational load is analyzed in the respect of optimal control theory and revised so that its estimation accuracy is enhanced. If we apply this proposed approach to optic motion capture systems, we can obtain additional advantages that motion capture can be performed without any markers, and with low cost in the respect of equipments and spaces.

• Journal of Broadcast Engineering
• /
• v.23 no.3
• /
• pp.409-420
• /
• 2018

In this paper, we propose a Motion Vector-based Particle Filter(MVPF) for object tracking on bitstreams and a object tracking system using the MVPF. The MVPF uses motion vectors to both the transition and the observation models of a general particle filter to improve the accuracy while maintaining the number of particles. In the proposed object tracking system, the state of the target object can be predicted using the histogram of motion vectors extracted from the bitstream. In terms of precision, F-measure and IOU(Intersection Of Union), the proposed method is about 30%, 17%, and 17% better on average, respectively, in MPEG test sequences and VOT2013 sequences. Furthermore, When the tracking results are displayed in box form for subjective performance evaluation, the proposed method can track moving objects more robust than the conventional methods in all test sequences.

• The KIPS Transactions:PartB
• /
• v.8B no.5
• /
• pp.447-454
• /
• 2001

The automatic document classification is a method that assigns unlabeled documents to the existing classes. The automatic document classification can be applied to a classification of news group articles, a classification of web documents, showing more precise results of Information Retrieval using a learning of users. In this paper, we use the weighted Bayesian classifier that weights with keywords of a document to improve the classification accuracy. If the system cant classify a document properly because of the lack of the number of words as the feature of a document, it uses relevance word cluster to supplement the feature of a document. The clusters are made by the automatic word clustering from the corpus. As the result, the proposed system outperformed existing classification system in the classification accuracy on Korean documents.

• The Journal of the Korea Contents Association
• /
• v.13 no.2
• /
• pp.52-61
• /
• 2013

In wireless sensor networks, the geographical positioning scheme is one of core technologies for sensor applications such as disaster monitoring and environment monitoring. For this reason, studies on range-free positioning schemes have been actively progressing. The density probability scheme based on central limit theorem and normal distribution was proposed to improve the location accuracy in non-uniform sensor network environments. The density probability scheme measures the final positions of unknown nodes by estimating distance through the sensor node communication. However, it has a problem that all of the neighboring nodes have the same 1-hop distance. In this paper, we propose an efficient sensor positioning scheme that overcomes this problem. The proposed scheme performs the second positioning step through the sensing range control after estimating the 1-hop distance of each node in order to minimize the estimation error. Our experimental results show that our proposed scheme improves the accuracy of sensor positioning by about 9% over the density probability scheme and by about 48% over the DV-HOP scheme.

• Journal of the Korean Society of Surveying, Geodesy, Photogrammetry and Cartography
• /
• v.9 no.2
• /
• pp.37-48
• /
• 1991

In this study, conventional network adjustment and combined network adjustment methods for single network adjustment methods for single network and centric combination network were compared by the analysis of root mean square error and standard error ellipse of observed points. It can be concluded from this study that for conventional surveying methods, the accuracy is in theorder of trilateration, traverse and triangulation, and for the case of combined surveying method, the accuracy is in the order of multilateration surveying, combined traverse and combined triangulation-trilateration surveying. And when establishing new control points, the accuracy can be improved by increasing redundant observations of centric combination network instead of using the single network. Also, in case of combined traverse surveying by adding observable laterals, accuracy level of trilateration could be achieved, and it was found that traverse is effective for large areas where sighting is easy, and combined traverse surveying is effective for urban areas where sighting is difficult.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.39A no.11
• /
• pp.645-651
• /
• 2014

For blind equalization, we propose a method of updating an equalizer, which generates an error from selectively applying a transmitted symbol constellation and that of induced equivalently from the transmitted symbol constellation and updates the equalizer by using this error. The proposed method, by selectively using the symbol constellation effective for improvement of symbol estimation accuracy and that of effective for improvement of error performance, showed that it is possible to improve the error performance at the same time to open the eye diagram of equalizer output quickly. As a criterion applying the symbol constellation, we used the dispersion of symbol points of equalizer output. In addition, to increase the accuracy of updating an equalizer the error was controlled by using current and previous dispersions. By simulation, under multipath channel with additive noise, we verified the equalization performance of the proposed method for 64-QAM.

• Journal of the Korea Society of Computer and Information
• /
• v.25 no.12
• /
• pp.203-210
• /
• 2020

Memory-based collaborative filtering is one of the representative types of the recommender system, but it suffers from the inherent problem of data sparsity. Although many works have been devoted to solving this problem, there is still a request for more systematic approaches to the problem. This study exploits distribution of user ratings given to items for computing similarity. All user ratings are utilized in the proposed method, compared to previous ones which use ratings for only common items between users. Moreover, for similarity computation, it takes a global view of ratings for items by reflecting other users' ratings for that item. Performance is evaluated through experiments and compared to that of other relevant methods. The results reveal that the proposed demonstrates superior performance in prediction and rank accuracies. This improvement in prediction accuracy is as high as 2.6 times more than that achieved by the state-of-the-art method over the traditional similarity measures.

• Proceedings of the Korea Water Resources Association Conference
• /
• 2023.05a
• /
• pp.458-458
• /
• 2023

본 연구에서는 위성영상 활용 지능형 재난관측·감시 기술 개발을 목적으로 위성영상과 멀티소스(CCTV, 항공영상, 공공DB 등)와의 연계·융합을 통해 재난상황관리의 정확도 향상과 위성영상 활용성 제고 방안을 제시하고자 하였다. 위성영상 수집·배포시스템으로부터 전달되는 위성영상과 멀티소스의 연계 융합을 통한 재난상황정보의 표출을 목적으로 상황판연계 표출시스템 가동 절차와 위성영상 수집을 통한 위험탐지 알고리즘과의 연계를 위해 재난상황업무 기반 시스템 가동절차를 수립하고, 위기관리표준 매뉴얼 상 상황업무절차를 적용해 예비설계를 진행하였다. 상황실 실무자 설문을 통해 작성된 시스템 요구사항과 규격서를 기반으로 상황업무절차를 적용해 먼저업무시스템 설계를 진행하였다. 평시에는 GIS통합상황판에서 관리됨을 전제로 위성영상 수집에 대한국가적 예산 투입 측면을 고려해 중대본 설치가 필요한 대형재난 발생상황을 가정하여 상황판연계·표출시스템의 가동되도록 설계하였다. 또한, 위성영상 분석을 통한 피해위험도와 재난이력통계 등 멀티소스와 중첩한 결과를 실시간으로 표출함에 따라 상황실근무자는 재난확산 여부를 판단하고, NDMS를 통해 재난상황을 전파할 수 있도록 설계하였다. 상황판연계 표출시스템의 원활한 데이터 입/출력을 위해 재난유형 및 분석단계별 클래스 정의, 유스케이스 ID(요구기능)와 1:1 또는 1:n매칭을 수행하여 재난유형 및 분석단계별 클래스를 정의하였다. 정의된 클래스는 유스케이스인 요구기능과 매칭을 수행하였고, 시스템 가동절차 중 피해위험도분석, 재난이력통계, 중첩결과표출, NDMS 상황전파에 대한 상황업무절차를 기반으로 산불·홍수·산사태·대설·태풍 총 5종의재난별 시퀀스를 설계하였다. 마지막으로 화면정의서와 UI/UX설계서를 기반으로 Figma를 통해 시스템구동화면을 사전에 모의하였다. 향후, 진행되는 연구에서는 위성영상과 멀티소스를 연계한 화면을 실체화하여 더욱 정확한 재난상황관리가 가능하도록 NDMS 연계 상황판 표출 시스템을 개발하고자 한다.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.18 no.5
• /
• pp.999-1008
• /
• 2023

Early prediction of chronic diseases such as diabetes is an important issue, and improving the accuracy of diabetes prediction is especially important. Various machine learning and deep learning-based methodologies are being introduced for diabetes prediction, but these technologies require large amounts of data for better performance than other methodologies, and the learning cost is high due to complex data models. In this study, we aim to verify the claim that DNN using the pima dataset and k-fold cross-validation reduces the efficiency of diabetes diagnosis models. Machine learning classification methods such as decision trees, SVM, random forests, logistic regression, KNN, and various ensemble techniques were used to determine which algorithm produces the best prediction results. After training and testing all classification models, the proposed system provided the best results on XGBoost classifier with ADASYN method, with accuracy of 81%, F1 coefficient of 0.81, and AUC of 0.84. Additionally, a domain adaptation method was implemented to demonstrate the versatility of the proposed system. An explainable AI approach using the LIME and SHAP frameworks was implemented to understand how the model predicts the final outcome.