• Journal of KIISE:Software and Applications
• /
• v.30 no.3_4
• /
• pp.358-371
• /
• 2003

Automatic word spacing is a process of deciding correct boundaries between words in a sentence including spacing errors. It is very important to increase the readability and to communicate the accurate meaning of text to the reader. The previous statistical approaches for automatic word spacing do not consider the previous spacing state, and thus can not help estimating inaccurate probabilities. In this paper, we propose two statistical word spacing models which can solve the problem of the previous statistical approaches. The proposed models are based on the observation that the automatic word spacing is regarded as a classification problem such as the POS tagging. The models can consider broader context and estimate more accurate probabilities by generalizing hidden Markov models. We have experimented the proposed models under a wide range of experimental conditions in order to compare them with the current state of the art, and also provided detailed error analysis of our models. The experimental results show that the proposed models have a syllable-unit accuracy of 98.33% and Eojeol-unit precision of 93.06% by the evaluation method considering compound nouns.

• The KIPS Transactions:PartB
• /
• v.16B no.5
• /
• pp.367-376
• /
• 2009

In this paper, we propose a flame detection method using Haar wavelet and moving averages in outdoor infrared video sequences. Our proposed method is composed of three steps which are Haar wavelet decomposition, flame candidates detection, and their tracking and flame classification. In Haar wavelet decomposition, each frame is decomposed into 4 sub- images(LL, LH, HL, HH), and also computed high frequency energy components using LH, HL, and HH. In flame candidates detection, we compute a binary image by thresholding in LL sub-image and apply morphology operations to the binary image to remove noises. After finding initial boundaries, final candidate regions are extracted using expanding initial boundary regions to their neighborhoods. In tracking and flame classification, features of region size and high frequency energy are calculated from candidate regions and tracked using queues, and we classify whether the tracked regions are flames by temporal changes of moving averages.

• Journal of the Korean Society of Environmental Restoration Technology
• /
• v.3 no.3
• /
• pp.20-28
• /
• 2000

The objective of this study is to establish a methodology of bioregional approach for coastal area management as a basis for planning and design. Focusing on the bioregional approach, this study reviewed currently prevailing approaches such as watershed approach and ecological unit approach for planning and management purposes. This research placed its geographical focus on the landward watershed of the Bay of Hampyong located in Chonnam Province, dealing efficiently with shortcomings of existing researches which mainly covered seaward tidal flats without considering outside effects. The main methods of the study are classified into indoor computerized map analysis and field work. For computer analysis, printed maps and digital maps have been analysed, and GIS techniques have been utilized for its synthesis and finalizations. Field work included on-site landscape analysis and verification of a tentative place unit boundary. As a practical step, criteria for classifying bioregion were presented and the selected criteria included : topography & water ways ; roads & administrative boundaries ; habitat types ; and visual enclosure. First, based on the data of topography and water ways, broad classification work was performed and corrections were made based on data drawn out from other criteria. A tentative place unit map was drawn and revised through field visits. This study encompassed an initial but integral part for bioregional approach in landward watershed management of a coastal area. As results of the study, the necessity and efficiency of bioregional approach which considers environmental and cultural components systematically have been presented.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.13 no.6
• /
• pp.679-685
• /
• 2003

The cores and deltas of fingerprints designate the parts where the flow of the ridges change radically. Observations on the change of the orientation of the ridges around the cores and deltas enable us to guess the location of the cores and deltas. According]y clustering the orientation flowing to the same direction after doing research on the orientation of the ridges on the whole makes us see that the cores and deltas are shaping around the boundaries of the clustering area. It is also observed that The patterns of clustering of the orientation of the ridges classified as Arch, Tented Arch, Left loop, Right Loop and Whorl have its own characteristics respectively. In this paper the method of classifying the fingerprints effectively is proposed and proved its effectiveness by using the clustering of the orientation of the ridges, finding the cores of the fingerprints which don't secure the deltas.

• Journal of the Society of Naval Architects of Korea
• /
• v.42 no.5 s.143
• /
• pp.479-486
• /
• 2005

The final goal of shift alignment design is that the bearing reaction forces or mean pressures are within design boundaries for various service conditions of a ship. However, it is found that calculated bearing load can be substantially variable according to the locations of the effective support points of after sterntube bearing which are determined by simple calculation or assumption suggested by classification societies. A new analysis method for shaft alignment calculation is introduced in order to resolve these problems. Key concept of the new method is featured by adopting both nonlinear elastic and multi-support elements to simulate a bearing support Hertz contact theory is basically applied for nonlinear elastic stiffness calculation instead of the projected area method suggested by most of classification societies. Three loading conditions according to the bearing offset and the hydrodynamic moment and twelve models according to the locations of the effective support points of sterntube bearings are prepared to carry out quantitative verifications for an actual shafting system of 8000 TEU class container vessel. It is found that there is relatively large difference between assumed and calculated effective support points.

• Geophysics and Geophysical Exploration
• /
• v.23 no.3
• /
• pp.131-140
• /
• 2020

In training a classification model using machine learning, the acquisition of training data is a very important stage, because the amount and quality of the training data greatly influence the model performance. However, when the cost of obtaining data is so high that it is difficult to build ideal training data, the number of samples for each class may be acquired very differently, and a serious data-imbalance problem can occur. If such a problem occurs in the training data, all classes are not trained equally, and classes containing relatively few data will have significantly lower recall values. Additionally, the reliability of evaluation indices such as accuracy and precision will be reduced. Therefore, this study sought to overcome the problem of data imbalance in two stages. First, we introduced weighted accuracy and weighted precision as new evaluation indices that can take into account a data-imbalance ratio by modifying conventional measures of accuracy and precision. Next, oversampling was performed to balance weighted precision and recall among classes. We verified the algorithm by applying it to the problem of facies classification. As a result, the imbalance between majority and minority classes was greatly mitigated, and the boundaries between classes could be more clearly identified.

• Transactions of the Korean Society for Noise and Vibration Engineering
• /
• v.15 no.1 s.94
• /
• pp.53-62
• /
• 2005

This study presents the NDE (non-destructive evaluation) technique for detecting the loosened bolts on joint steel structures on the basis of TOF (time of flight) and amplitudes of Lamb waves. Probabilistic neural network (PNN) technique which is an effective tool for pattern classification problem was applied to the damage estimation using PZT induced Lamb waves. Two kinds of damages were introduced by dominant damages (DD) which mean loosened bolts within the Lamb waves beam width and minor damages (MD) which mean loosened bolts out of the Lamb waves beam width. They were investigated for the establishment of the optimal decision boundaries which divide each damage class's region including the intact class. In this study, the applicability of the probabilistic neural networks was identified through the test results for the damage cases within and out of wave beam path. It has been found that the present methods are very efficient and reasonable in predicting the loosened bolts on the joint steel structures probabilistically.

• Proceedings of the KSRS Conference
• /
• 1999.11a
• /
• pp.343-348
• /
• 1999

Delineation of watersheds is one of the most basic steps for water resource management and National Park management. The purpose of this study is to investigate how to utilize Thematic Mappers scenes to compare watersheds created by running a model with those produced by digitizing topographic maps of Keum River basin. A methodology is designed and tested using Geographic Information Systems (GIS) and remote sensing to map areas with various thematic maps. A CAD data on watersheds from the Decision Support system for Water Quality is converted into GIS format. The digital elevation model with 100-meter resolution is used to create watershed by cumulative watershed method. TM scenes are also classified by new procedures including stacking method, NDVI, NDWI, and unsupervised classification methods. To evaluate the relative correctness Kyerongsan National Park was studied intensively whose area was divided into 6 watersheds in both cases. The boundaries of watershed from the model are less correct than those of the topographic maps. This result shows that automated watershed creating system needs higher-resolution digital elevation model than 100-meters.

• Journal of Korean Institute of Industrial Engineers
• /
• v.28 no.2
• /
• pp.147-161
• /
• 2002

In novelty detection, one attempts to discriminate abnormal patterns from normal ones. Novelty detection is quite difficult since, unlike usual two class classification problems, only normal patterns are available for training. Auto-Associative Multi-Layer Perceptron (AAMLP) has been shown to provide a good performance based upon the property that novel patterns usually have larger auto-associative errors. In this paper, we give a mathematical analysis of 2-layer AAMLP's output characteristics and empirical results of 2-layer and 4-layer AAMLPs. Various activation functions such as linear, saturated linear and sigmoid are compared. The 2-layer AAMLPs cannot identify non-linear boundaries while the 4-layer ones can. When the data distribution is multi-modal, then an ensemble of AAMLPs, each of which is trained with pre-clustered data is required. This paper contributes to understanding of AAMLP networks and leads to practical recommendations regarding its use.

• Interdisciplinary Bio Central
• /
• v.4 no.4
• /
• pp.14.1-14.6
• /
• 2012

Splice site prediction in DNA sequence is a basic search problem for finding exon/intron and intron/exon boundaries. Removing introns and then joining the exons together forms the mRNA sequence. These sequences are the input of the translation process. It is a necessary step in the central dogma of molecular biology. The main task of splice site prediction is to find out the exact GT and AG ended sequences. Then it identifies the true and false GT and AG ended sequences among those candidate sequences. In this paper, we survey research works on splice site prediction based on support vector machine (SVM). The basic difference between these research works is nucleotide encoding technique and SVM kernel selection. Some methods encode the DNA sequence in a sparse way whereas others encode in a probabilistic manner. The encoded sequences serve as input of SVM. The task of SVM is to classify them using its learning model. The accuracy of classification largely depends on the proper kernel selection for sequence data as well as a selection of kernel parameter. We observe each encoding technique and classify them according to their similarity. Then we discuss about kernel and their parameter selection. Our survey paper provides a basic understanding of encoding approaches and proper kernel selection of SVM for splice site prediction.