• Journal of the Korean Institute of Intelligent Systems
• /
• v.22 no.3
• /
• pp.319-327
• /
• 2012

In this paper, we present a sign language recognition model which does not use any wearable devices for object tracking. The system design issues and implementation issues such as data representation, feature extraction and pattern classification methods are discussed. The proposed data representation method for sign language patterns is robust for spatio-temporal variances of feature points. We present a feature extraction technique which can improve the computation speed by reducing the amount of feature data. A neural network model which is capable of incremental learning is described and the behaviors and learning algorithm of the model are introduced. We have defined a measure which reflects the relevance between the feature values and the pattern classes. The measure makes it possible to select more effective features without any degradation of performance. Through the experiments using six types of sign language patterns, the proposed model is evaluated empirically.

• Journal of Plant Biotechnology
• /
• v.45 no.4
• /
• pp.306-314
• /
• 2018

Korean ginseng (Panax ginseng) has long been cultivated as an important economic medicinal plant. Owing to the seasonal and long-term agricultural cultivation methods of Korean ginseng, they are always vulnerable to various environmental stress conditions. ABSCISIC ACID (ABA) is an essential plant hormone associated with seed development and diverse abiotic stress responses including drought, cold and salinity stress. By modulating ABA responses, plants can regulate their immune responses and growth patterns to increase their ability to tolerate stress. With recent advances in genome sequencing technology, we first reported the functional features of genes related to canonical ABA signaling pathway in P. ginseng genome. Based on the protein sequences and functional genomic analysis of Arabidopsis thaliana, the ABA related genes were successfully identified. Our functional genomic characterizations clearly showed that the ABA signaling related genes consisting the ABA receptor proteins (PgPYLs), kinase family (PgSnRKs) and transcription factors (PgABFs, PgABI3s and PgABI5s) were evolutionary conserved in the P. ginseng genome. We confirmed that overexpressing ABA related genes of P. ginseng completely restored the ABA responses and stress tolerance in ABA defective Arabidopsis mutants. Finally, tissue and age specific spatio-temporal expression patterns of the identified ABA-related genes in P. ginseng tissues were also classified using various available RNA sequencing data. This study provides ABA signal transduction related genes and their functional genomic information related to the growth and development of Korean ginseng. Additionally, the results of this study could be useful in the breeding or artificial selection of ginseng which is resistant to various stresses.

• Journal of the Korea Society of Computer and Information
• /
• v.26 no.4
• /
• pp.11-19
• /
• 2021

An abnormal object refers to a person, an object, or a mechanical device that performs abnormal and unusual behavior and needs observation or supervision. In order to detect this through artificial intelligence algorithm without continuous human intervention, a method of observing the specificity of temporal features using optical flow technique is widely used. In this study, an abnormal situation is identified by learning an algorithm that translates an input image frame to an optical flow image using a Generative Adversarial Network (GAN). In particular, we propose a technique that improves the pre-processing process to exclude unnecessary outliers and the post-processing process to increase the accuracy of identification in the test dataset after learning to improve the performance of the model's abnormal behavior identification. UCSD Pedestrian and UMN Unusual Crowd Activity were used as training datasets to detect abnormal behavior. For the proposed method, the frame-level AUC 0.9450 and EER 0.1317 were shown in the UCSD Ped2 dataset, which shows performance improvement compared to the models in the previous studies.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.23 no.3
• /
• pp.145-151
• /
• 2023

3D-CNN is one of the deep learning techniques for learning time series data. Such three-dimensional learning can generate many parameters, so that high-performance machine learning is required or can have a large impact on the learning rate. When learning dynamic hand-gestures in spatiotemporal domain, it is necessary for the improvement of the efficiency of dynamic hand-gesture learning with 3D-CNN to find the optimal conditions of input video data by analyzing the learning accuracy according to the spatiotemporal change of input video data without structural change of the 3D-CNN model. First, the time ratio between dynamic hand-gesture actions is adjusted by setting the learning interval of image frames in the dynamic hand-gesture video data. Second, through 2D cross-correlation analysis between classes, similarity between image frames of input video data is measured and normalized to obtain an average value between frames and analyze learning accuracy. Based on this analysis, this work proposed two methods to effectively select input video data for 3D-CNN deep learning of dynamic hand-gestures. Experimental results showed that the learning interval of image data frames and the similarity of image frames between classes can affect the accuracy of the learning model.

• Journal of Broadcast Engineering
• /
• v.23 no.3
• /
• pp.351-360
• /
• 2018

Human emotion recognition is a research topic that is receiving continuous attention in computer vision and artificial intelligence domains. This paper proposes a method for classifying human emotions through multiple neural networks based on multi-modal signals which consist of image, landmark, and audio in a wild environment. The proposed method has the following features. First, the learning performance of the image-based network is greatly improved by employing both multi-task learning and semi-supervised learning using the spatio-temporal characteristic of videos. Second, a model for converting 1-dimensional (1D) landmark information of face into two-dimensional (2D) images, is newly proposed, and a CNN-LSTM network based on the model is proposed for better emotion recognition. Third, based on an observation that audio signals are often very effective for specific emotions, we propose an audio deep learning mechanism robust to the specific emotions. Finally, so-called emotion adaptive fusion is applied to enable synergy of multiple networks. The proposed network improves emotion classification performance by appropriately integrating existing supervised learning and semi-supervised learning networks. In the fifth attempt on the given test set in the EmotiW2017 challenge, the proposed method achieved a classification accuracy of 57.12%.

• Journal of the Korean Association of Geographic Information Studies
• /
• v.10 no.4
• /
• pp.60-74
• /
• 2007

For quantitative analysis of spatio-temporal distribution pattern on marine environment, spatial autocorrelation statistics on the both global and local aspects was applied to the observed data obtained from Gwangyang Bay in South Sea of Korea. Global indexes such as Moran's I and General G were used for understanding environmental distribution pattern in the whole study area. LISAs (local indicators of spatial association) such as Moran's I ($I_i$) and $G_i{^*}$ were considered to find similarity between a target feature and its neighborhood features and to detect hot spot and/or cold spot. Additionally, the significance test on clustered patterns by Z-scores was carried out. Statistical results showed variations of spatial patterns quantitatively in the whole year. Then all of general water quality, nutrients, chlorophyll-a and phytoplankton had strong clustered pattern in summer. When global indexes showed strong clustered pattern, the front region with a negative $I_i$ which means a strong spatial variation was observed. Also, when global indexes showed random pattern, hot spot and/or cold spot were/was found in the small local region with a local index $G_i{^*}$. Therefore, global indexes were useful for observing the strength and time series variations of clustered patterns in the whole study area, and local indexes were useful for tracing the location of hot spot and/or cold spot. Quantification of both spatial distribution pattern and clustering characteristics may play an important role to understand marine environment in depth and to find the reasons for spatial pattern.

• The Bulletin of The Korean Astronomical Society
• /
• v.37 no.2
• /
• pp.113.1-113.1
• /
• 2012

The characteristics of Doppler shifts in a quiet region of the Sun are investigated by comparing between the $H{\alpha}$ line and the Caii infrared line at 854.2 nm. A small area of $16^{\prime\prime}{\times}40^{\prime\prime}$ was observed for about half an hour with the Fast Imaging Solar Spectrograph (FISS) of the 1.6 meter New Solar Telescope (NST) at Big Bear Solar Observatory. The observed area contains a network region and an internetwork region, and identified in the network region are $H{\alpha}$ fibrils, Caii fibrils and bright points. We infer the Doppler velocity from each line profile at a point with the lambdameter method as a function of half wavelength separation ${\Delta}{\lambda}$. It is confirmed that the bisector of the spatially-averaged Caii line profile has an inverse C-shape of with a significant peak redshift of +1.8 km/s. In contrast, the bisector of the spatially-averaged $H{\alpha}$ line profile has a different shape; it is almost vertically straight or, if not, has a C-shape with a small peak blueshift of -0.5 km/s. In both the lines, the bisectors of bright network points are much different from those of other features in that they are significantly redshifted not only at the line centers, but also at the wings. We also find that the spatio-temporal fluctuation of Doppler shift inferred from the Caii line is correlated with those of the $H{\alpha}$ line. The strongest correlation occurs in the internework region, and when the inner wings rather than the line centers are used to determine Doppler shift. In this region, the RMS value of Doppler shift fluctuation is the largest at the line center, and monotonically decreases with ${\Delta}{\lambda}$. We discuss the physical implications of our results on the formation of the $H{\alpha}$ line and Caii 854.2 nm line in the quiet region chromosphere.

• The Sea:JOURNAL OF THE KOREAN SOCIETY OF OCEANOGRAPHY
• /
• v.27 no.2
• /
• pp.71-86
• /
• 2022

The water quality index (WQI) has been widely used to evaluate marine water quality. The WQI in Korea is categorized into five classes by marine environmental standards. But, the WQI calculation on huge datasets is a very complex and time-consuming process. In this regard, the current study proposed machine learning (ML) based models to predict WQI class by using water quality datasets. Sihwa Lake, one of specially-managed coastal zone, was selected as a modeling site. In this study, adaptive boosting (AdaBoost) and tree-based pipeline optimization (TPOT) algorithms were used to train models and each model performance was evaluated by metrics (accuracy, precision, F1, and Log loss) on classification. Before training, the feature importance and sensitivity analysis were conducted to find out the best input combination for each algorithm. The results proved that the bottom dissolved oxygen (DO_Bot) was the most important variable affecting model performance. Conversely, surface dissolved inorganic nitrogen (DIN_Sur) and dissolved inorganic phosphorus (DIP_Sur) had weaker effects on the prediction of WQI class. In addition, the performance varied over features including stations, seasons, and WQI classes by comparing spatio-temporal and class sensitivities of each best model. In conclusion, the modeling results showed that the TPOT algorithm has better performance rather than the AdaBoost algorithm without considering feature selection. Moreover, the WQI class for unknown water quality datasets could be surely predicted using the TPOT model trained with satisfactory training datasets.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.19 no.1
• /
• pp.137-142
• /
• 2024

Recently, deep learning technology has become those methods as de facto standards in the area of medical data representation. But, deep learning inherently requires a large amount of training data, which poses a challenge for its direct application in the medical field where acquiring large-scale data is not straightforward. Additionally, brain signal modalities also suffer from these problems owing to the high variability. Research has focused on designing deep neural network structures capable of effectively extracting spectro-spatio-temporal characteristics of brain signals, or employing self-supervised learning methods to pre-learn the neurophysiological features of brain signals. This paper analyzes methodologies used to handle small-scale data in emerging fields such as brain-computer interfaces and brain signal-based state prediction, presenting future directions for these technologies. At first, this paper examines deep neural network structures for representing brain signals, then analyzes self-supervised learning methodologies aimed at efficiently learning the characteristics of brain signals. Finally, the paper discusses key insights and future directions for deep learning-based brain signal analysis.

• Journal of Soil and Groundwater Environment
• /
• v.12 no.5
• /
• pp.19-32
• /
• 2007

In Korea, there have been various methods of estimating groundwater recharge which generally can be subdivided into three types: baseflow separation method by means of groundwater recession curve, water budget analysis based on lumped conceptual model in watershed, and water table fluctuation method (WTF) by using the data from groundwater monitoring wells. However, groundwater recharge rate shows the spatial-temporal variability due to climatic condition, land use and hydrogeological heterogeneity, so these methods have various limits to deal with these characteristics. To overcome these limitations, we present a new method of estimating recharge based on water balance components from the SWAT-MODFLOW which is an integrated surface-ground water model. Groundwater levels in the interest area close to the stream have dynamics similar to stream flow, whereas levels further upslope respond to precipitation with a delay. As these behaviours are related to the physical process of recharge, it is needed to account for the time delay in aquifer recharge once the water exits the soil profile to represent these features. In SWAT, a single linear reservoir storage module with an exponential decay weighting function is used to compute the recharge from soil to aquifer on a given day. However, this module has some limitations expressing recharge variation when the delay time is too long and transient recharge trend does not match to the groundwater table time series, the multi-reservoir storage routing module which represents more realistic time delay through vadose zone is newly suggested in this study. In this module, the parameter related to the delay time should be optimized by checking the correlation between simulated recharge and observed groundwater levels. The final step of this procedure is to compare simulated groundwater table with observed one as well as to compare simulated watershed runoff with observed one. This method is applied to Mihocheon watershed in Korea for the purpose of testing the procedure of proper estimation of spatio-temporal groundwater recharge distribution. As the newly suggested method of estimating recharge has the advantages of effectiveness of watershed model as well as the accuracy of WTF method, the estimated daily recharge rate would be an advanced quantity reflecting the heterogeneity of hydrogeology, climatic condition, land use as well as physical behaviour of water in soil layers and aquifers.