• The Transactions of The Korean Institute of Electrical Engineers
• /
• v.67 no.11
• /
• pp.1486-1495
• /
• 2018

This paper investigates the autonomous ground vehicle (AGV) localization filter design problem under GNSS-denied environments. It is assumed that the given landmarks do not have unique features due to the lack of a prior knowledge on them. For such case, the AGV may have difficulties in distinguishing the position measurement of the detected landmark from those of other landmarks with the same feature, hence the conventional localization filters are not applicable. To resolve this technical issue, the localization filter design problem is formulated as a special form of the data association determining whether the detected feature is actually originated from which landmark. The measurement hypotheses generated by landmarks with the same feature are evaluated by the nearest neighbor data association scheme to reduce the computational burden. The position measurement corresponding to the landmark with the most probable hypothesis is used for localization filter. Through the experiments in real-driving condition, it is shown that the proposed method provides satisfactory localization performance in spite of using non-unique landmarks.

• Journal of the Korean Society of Surveying, Geodesy, Photogrammetry and Cartography
• /
• v.25 no.6_1
• /
• pp.529-536
• /
• 2007

Registration between image and object space is a fundamental step in photogrammetry and computer vision. Along with rapid development of sensors - multi/hyper spectral sensor, laser scanning sensor, radar sensor etc., the needs for registration between different sensors are ever increasing. There are two important considerations on different sensor registration. They are sensor invariant feature extraction and correspondence between them. Since point to point correspondence does not exist in image and laser scanning data, it is necessary to have higher entities for extraction and correspondence. This leads to modify first, existing mathematical and geometrical model which was suitable for point measurement to line measurements, second, matching scheme. In this research, linear feature is selected for sensor invariant features and matching entity. Linear features are incorporated into mathematical equation in the form of extended collinearity equation for registration problem known as photo resection which calculates exterior orientation parameters. The other emphasis is on the scheme of finding matched entities in the aide of RANSAC (RANdom SAmple Consensus) in the absence of correspondences. To relieve computational load which is a common problem in sampling theorem, deterministic sampling technique and selecting 4 line features from 4 sectors are applied.

• The Journal of Korea Robotics Society
• /
• v.13 no.1
• /
• pp.16-25
• /
• 2018

The talking head (TH) indicates an utterance face animation generated based on text and voice input. In this paper, we propose the generation method of TH with facial expression and intonation by speech input only. The problem of generating TH from speech can be regarded as a regression problem from the acoustic feature sequence to the facial code sequence which is a low dimensional vector representation that can efficiently encode and decode a face image. This regression was modeled by bidirectional RNN and trained by using SAVEE database of the front utterance face animation database as training data. The proposed method is able to generate TH with facial expression and intonation TH by using acoustic features such as MFCC, dynamic elements of MFCC, energy, and F0. According to the experiments, the configuration of the BLSTM layer of the first and second layers of bidirectional RNN was able to predict the face code best. For the evaluation, a questionnaire survey was conducted for 62 persons who watched TH animations, generated by the proposed method and the previous method. As a result, 77% of the respondents answered that the proposed method generated TH, which matches well with the speech.

• Journal of Korea Society of Digital Industry and Information Management
• /
• v.11 no.4
• /
• pp.47-60
• /
• 2015

In this paper, we propose a simple and useful method using a port role to visualize the network attacks. The port role defines the behavior of the port from the source and destination port number of network session. Based on the port role, the port provides the brief security features of each node as an attacker, a victim, a server, and a normal host. We have automatically classified and identified the type of node based on the port role and security features. We detected and visualized the network attacks using these features of the node by the port role. In addition, we are intended to solve the problems with existing visualization technologies which are the reflection problem caused an undirected network session and the problem caused decreasing of distinct appearance when occurs a large amount of the sessions. The proposed method monitors anomalies occurring in an entire network and displays detailed information of the attacker, victim, server, and hosts. In addition, by providing a categorized analysis of network attacks, this method can more precisely detect and distinguish them from normal sessions.

• Journal of the Korea Society for Simulation
• /
• v.20 no.4
• /
• pp.13-20
• /
• 2011

The face recognition has been used in a variety fields, such as identification and security. The procedure of the face recognition is as follows; extracting face features of face images, learning the extracted face features, and selecting some features among all extracted face features. The selected features have discrimination and are used for face recognition. However, there are numerous face features extracted from face images. If a face recognition system uses all extracted features, a high computing time is required for learning face features and the efficiency of computing resources decreases. To solve this problem, many researchers have proposed various Boosting methods, which improve the performance of learning algorithms. Mutual-Boost is the typical Boosting method and efficiently selects face features by using mutual information between two features. In this paper, we propose a GroupMutual-Boost method for improving Mutual-Boost. Our proposed method can shorten the time required for learning and recognizing face features and use computing resources more effectively since the method does not learn individual features but a feature group.

• ETRI Journal
• /
• v.35 no.5
• /
• pp.900-910
• /
• 2013

Small-screen mobile terminals have difficulty accessing existing Web resources designed for large-screen devices. This paper presents an adaptive transformation method based on webpage semantic features to solve this problem. According to the text density and link density features of the webpages, the webpages are divided into two types: index and content. Our method uses an index-based webpage transformation algorithm and a content-based webpage transformation algorithm. Experiment results demonstrate that our adaptive transformation method is not dependent on specific software and webpage templates, and it is capable of enhancing Web content adaptation on small-screen terminals.

• The Journal of the Acoustical Society of Korea
• /
• v.21 no.4E
• /
• pp.156-163
• /
• 2002

We apply independent component analysis (ICA) for extracting an optimal basis to the problem of finding efficient features for representing speech signals of a given speaker The speech segments are assumed to be generated by a linear combination of the basis functions, thus the distribution of speech segments of a speaker is modeled by adapting the basis functions so that each source component is statistically independent. The learned basis functions are oriented and localized in both space and frequency, bearing a resemblance to Gabor wavelets. These features are speaker dependent characteristics and to assess their efficiency we performed speaker identification experiments and compared our results with the conventional Fourier-basis. Our results show that the proposed method is more efficient than the conventional Fourier-based features in that they can obtain a higher speaker identification rate.

• 한국HCI학회:학술대회논문집
• /
• 2008.02a
• /
• pp.504-509
• /
• 2008

Support Vector Machine (SVM) is one of powerful learning machine and has been applied to varying task with generally acceptable performance. The success of SVM for classification tasks in one domain is affected by features which represent the instance of specific class. Given the representative and discriminative features, SVM learning will give good generalization and consequently we can obtain good classifier. In this paper, we will assess the problem of feature choices for human detection tasks and measure the performance of each feature. Here we will consider HOG-family feature. As a natural extension of SVM, we combine SVM with Principal Component Analysis (PCA) to reduce dimension of features while retaining most of discriminative feature vectors.

• Journal of the Korea Institute of Military Science and Technology
• /
• v.15 no.3
• /
• pp.298-305
• /
• 2012

In this paper, we propose an integrated algorithm for scene recognition, which has been a challenging computer vision problem, with application to mobile robot localization. The proposed scene recognition method utilizes SIFT and visual words as local-level features and GIST as a global-level feature. As local-level and global-level features complement each other, it results in improved performance for scene recognition. This improved algorithm is of low computational complexity and robust to image distortions.

• The Journal of Information Technology and Database
• /
• v.8 no.2
• /
• pp.91-101
• /
• 2001

Previous research efforts on performance evaluation of multidimensional indexes typically have used synthetic data sets distributed uniformly or normally over multidimensional space. However, recent research research result has shown that these hinds of data sets hardly reflect the characteristics of multimedia database applications. In this paper, we discuss issues on generating high dimensional data and query sets for resolving the problem. We first identify the features of the data and query sets that are appropriate for fairly evaluating performances of multidimensional indexes, and then propose HDDQ_Gen(High-Dimensional Data and Query Generator) that satisfies such features. HDDQ_Gen supports the following features : (1) clustered distributions, (2) various object distributions in each cluster, (3) various cluster distributions, (4) various correlations among different dimensions, (5) query distributions depending on data distributions. Using these features, users are able to control tile distribution characteristics of data and query sets. Our contribution is fairly important in that HDDQ_Gen provides the benchmark environment evaluating multidimensional indexes correctly.