• Korean Journal of Applied Biomechanics
• /
• v.18 no.2
• /
• pp.95-104
• /
• 2008

The purpose of this study was to investigate kinematic coordination and control of lower segments in skill process. For the investigation, we examined the difference of resultant linear velocity of segments and angle vs angle graph. Novice subjects were 9 male middle school students who has never been experienced a taekwondo and expert subjects were 7 university taekwondo players. We analyzed kinematic variables of Dollyochagi motion through videographical analysis and the conclusion were as follows. 1. Examining the graph of novice subjects' maximal resultant linear velocity of the thigh, shank, and foot segment, as it gets closer to the end of the training, the maximal resultant linear velocity in each segment increased. Statistical analysis showed the following results; thigh segment caused the increase of speed, using the trunk segment's momentum in the latter term of learning, while the shank segment utilized the momentum of the adjacent proximal segment at the beginning of learning, and the foot segment in the middle of learning. 2. Until the point where the knee joint angle is minimum, as the novice group learn the skill, the flexion of knee and hip joints has changed into the form of coordination pattern in phase. On the other hand, the expert group showed continual coordination pattern in phase that the movement sequences were smooth. From the knee joint maximal flexion to impact timing, all novice and expert groups showed coordination pattern out of phase. 3. From the knee joint maximal flexion to impact timing, the ankle joint was fixed and the knee joint was extended to all the novice stages and expert subjects.

• The Transactions of The Korean Institute of Electrical Engineers
• /
• v.67 no.1
• /
• pp.114-123
• /
• 2018

This paper concerns a design and learning method of softmax function neural networks based on K-means clustering. The partial discharge data Information is preliminarily processed through simulation using an Epoxy Mica Coupling sensor and an internal Phase Resolved Partial Discharge Analysis algorithm. The obtained information is processed according to the characteristics of the pattern using a Motor Insulation Monitoring System program. At this time, the processed data are total 4 types that void discharge, corona discharge, surface discharge and slot discharge. The partial discharge data with high dimensional input variables are secondarily processed by principal component analysis method and reduced with keeping the characteristics of pattern as low dimensional input variables. And therefore, the pattern classifier processing speed exhibits improved effects. In addition, in the process of extracting the partial discharge data through the MIMS program, the magnitude of amplitude is divided into the maximum value and the average value, and two pattern characteristics are set and compared and analyzed. In the first half of the proposed partial discharge pattern classifier, the input and hidden layers are classified by using the K-means clustering method and the output of the hidden layer is obtained. In the latter part, the cross entropy error function is used for parameter learning between the hidden layer and the output layer. The final output layer is output as a normalized probability value between 0 and 1 using the softmax function. The advantage of using the softmax function is that it allows access and application of multiple class problems and stochastic interpretation. First of all, there is an advantage that one output value affects the remaining output value and its accompanying learning is accelerated. Also, to solve the overfitting problem, L2-normalization is applied. To prove the superiority of the proposed pattern classifier, we compare and analyze the classification rate with conventional radial basis function neural networks.

• Journal of IKEEE
• /
• v.24 no.2
• /
• pp.529-535
• /
• 2020

As the main technology of the 4th industrial revolution, immersive 360-degree video contents are drawing attention. The market size of immersive 360-degree video contents worldwide is projected to increase from $6.7 billion in 2018 to approximately $70 billion in 2020. However, most of the immersive 360-degree video contents are distributed through illegal distribution networks such as Webhard and Torrent, and the damage caused by illegal reproduction is increasing. Existing 2D video industry uses copyright filtering technology to prevent such illegal distribution. The technical difficulties dealing with immersive 360-degree videos arise in that they require ultra-high quality pictures and have the characteristics containing images captured by two or more cameras merged in one image, which results in the creation of distortion regions. There are also technical limitations such as an increase in the amount of feature point data due to the ultra-high definition and the processing speed requirement. These consideration makes it difficult to use the same 2D filtering technology for 360-degree videos. To solve this problem, this paper suggests a feature point extraction and identification technique that select object identification areas excluding regions with severe distortion, recognize objects using deep learning technology in the identification areas, extract feature points using the identified object information. Compared with the previously proposed method of extracting feature points using stitching area for immersive contents, the proposed technique shows excellent performance gain.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.21 no.3
• /
• pp.600-611
• /
• 1996

Recently, neural network-based speech recognition has been studied to utilize the adaptivity and learnability of neural network models. However, conventional neural network models have difficulty in the co-articulation processing and the boundary detection of similar phonmes of the Korean speech. Also, in case of using one phonotopic map, learning speed may dramatically increase and inaccuracies may be caused because homogeneous learning and recognition method should be applied for heterogenous data. Hence, in this paper, a neural net typewriter has been designed using a hierarchical self-organizing classifier(HSOC), and related algorithms are presented. This HSOC, during its learing stage, distributed phoneme data on hierarchically structured multiple phonotopic maps, using Kohonen's self-organizing feature maps(SOFM). Presented and experimented in this paper were the algorithms for deciding the number of maps, map sizes, the selection of phonemes and their placement per map, an approapriate learning and preprocessing method per map. If maps are divided according to a priorlinguistic knowledge, we would have difficulty in acquiring linguistic knowledge and how to alpply it(e.g., processing extended phonemes). Contrarily, our HSOC has an advantage that multiple phonotopic maps suitable for given input data are self-organizable. The resulting three korean phonotopic maps are optimally labelled and have their own optimal preprocessing schemes, and also confirm to the conventional linguistic knowledge.

• The Journal of The Korea Institute of Intelligent Transport Systems
• /
• v.16 no.4
• /
• pp.153-163
• /
• 2017

This research is based on the previous research that personally preferred safe distance, rotating angle and speed are differentiated. Thus, we use machine learning model for recognizing maneuvering modes trained per personal or per similar driving pattern groups, and we evaluate automatic driving according to maneuvering modes. By utilizing driving knowledge, we subdivided 8 kinds of longitudinal modes and 4 kinds of lateral modes, and by combining the longitudinal and lateral modes, we build 21 kinds of maneuvering modes. we train the labeled data set per time stamp through RNN, LSTM and Bi-LSTM models by the trips of drivers, which are supervised deep learning models, and evaluate the maneuvering modes of automatic driving for the test data set. The evaluation dataset is aggregated of living trips of 3,000 populations by VTTI in USA for 3 years and we use 1500 trips of 22 people and training, validation and test dataset ratio is 80%, 10% and 10%, respectively. For recognizing longitudinal 8 kinds of maneuvering modes, RNN achieves better accuracy compared to LSTM, Bi-LSTM. However, Bi-LSTM improves the accuracy in recognizing 21 kinds of longitudinal and lateral maneuvering modes in comparison with RNN and LSTM as 1.54% and 0.47%, respectively.

• Journal of the Korean Geographical Society
• /
• v.38 no.2
• /
• pp.275-291
• /
• 2003

It was essential to develop the DB-type teaching and teaming material for geography instruction using a method of ICT. The DB-type teaching and learning material was considered as a alternative in solving the problems of web-based geography instruction. Accordingly, in this study, the geography image DB program as developed, and based on this program the CD-ROM called GEO-DB, having the function of electronic dictionary of geography image for geography teaching and teaming was made. The GEO-DB was composed of 3,060 geography images collected by teachers and learners. The GEO-DB was made to be used simply by teachers and learners. Especially, the portfolio function was Included in the GEO-DB, and that was focused to the instructional system design of teacher and the self-directed teaming ability development of learner. Teachers and learners using this GEO-DB assessed that because the GEO-DB had the easiness of use, the speed of reference and the unlimitedness of extension, it could enlarge the possibility of using a method of In, and it could contribute to the development of geography teaming ability and the change of geography teaming attitude.

• Journal of the Korea Computer Industry Society
• /
• v.3 no.3
• /
• pp.331-340
• /
• 2002

The web world is getting so huge and untractable that without an intelligent information extractor we would get more and more helpless. Conventional web spidering techniques for general purpose search engine may be too slow for the specific search engines, which concentrate only on specific areas or keywords. In this paper a new model for improving web spidering capabilities is suggested and experimented. How to select adequate reference web pages from the initial web Page set relevant to a given specific area (or keywords) can be very important to reduce the spidering speed. Our reference web page selection method DOPS dynamically and orthogonally selects web pages, and it can also decide the appropriate number of reference pages, using a newly defined measure. Even for a very specific area, this method worked comparably well almost at the level of experts. If we consider that experts cannot work on a huge initial page set, and they still have difficulty in deciding the optimal number of the reference web pages, this method seems to be very promising. We also applied reinforcement learning to web environment, and DOPS-based reinforcement learning experiments shows that our method works quite favorably in terms of both the number of hyper links and time.

• Journal of Korean Tunnelling and Underground Space Association
• /
• v.20 no.6
• /
• pp.1161-1175
• /
• 2018

An unexpected event could be easily followed by a large secondary accident due to the limitation in sight of drivers in road tunnels. Therefore, a series of automated incident detection systems have been under operation, which, however, appear in very low detection rates due to very low image qualities on CCTVs in tunnels. In order to overcome that limit, deep learning based tunnel incident detection system was developed, which already showed high detection rates in November of 2017. However, since the object detection process could deal with only still images, moving direction and speed of moving vehicles could not be identified. Furthermore it was hard to detect stopping and reverse the status of moving vehicles. Therefore, apart from the object detection, an object tracking method has been introduced and combined with the detection algorithm to track the moving vehicles. Also, stopping-reverse discrimination algorithm was proposed, thereby implementing into the combined incident detection processes. Each performance on detection of stopping, reverse driving and fire incident state were evaluated with showing 100% detection rate. But the detection for 'person' object appears relatively low success rate to 78.5%. Nevertheless, it is believed that the enlarged richness of image big-data could dramatically enhance the detection capacity of the automatic incident detection system.

• The Journal of the Korea Contents Association
• /
• v.21 no.1
• /
• pp.531-540
• /
• 2021

As cyber attacks and crimes increase exponentially and hacking attacks become more intelligent and advanced, hacking attack methods and routes are evolving unpredictably and in real time. In order to reinforce the enemy's responsiveness, this study aims to propose a method for developing an artificial intelligence-based security control platform by building a next-generation security system using artificial intelligence to respond by self-learning, monitoring abnormal signs and blocking attacks.The artificial intelligence-based security control platform should be developed as the basis for data collection, data analysis, next-generation security system operation, and security system management. Big data base and control system, data collection step through external threat information, data analysis step of pre-processing and formalizing the collected data to perform positive/false detection and abnormal behavior analysis through deep learning-based algorithm, and analyzed data Through the operation of a security system of prevention, control, response, analysis, and organic circulation structure, the next generation security system to increase the scope and speed of handling new threats and to reinforce the identification of normal and abnormal behaviors, and management of the security threat response system, Harmful IP management, detection policy management, security business legal system management. Through this, we are trying to find a way to comprehensively analyze vast amounts of data and to respond preemptively in a short time.

• Journal of the Korean Society of Surveying, Geodesy, Photogrammetry and Cartography
• /
• v.40 no.5
• /
• pp.381-391
• /
• 2022

This paper is intended to find one of the prevailing deep learning models that are a type of AI (Artificial Intelligence) that helps rapidly detect damaged buildings where disasters occur. The models selected are SSD-512, RetinaNet, and YOLOv3 which are widely used in object detection in recent years. These models are based on one-stage detector networks that are suitable for rapid object detection. These are often used for object detection due to their advantages in structure and high speed but not for damaged building detection in disaster management. In this study, we first trained each of the algorithms on xBD dataset that provides the post-disaster imagery with damage classification labels. Next, the three models are quantitatively evaluated with the mAP(mean Average Precision) and the FPS (Frames Per Second). The mAP of YOLOv3 is recorded at 34.39%, and the FPS reached 46. The mAP of RetinaNet recorded 36.06%, which is 1.67% higher than YOLOv3, but the FPS is one-third of YOLOv3. SSD-512 received significantly lower values than the results of YOLOv3 on two quantitative indicators. In a disaster situation, a rapid and precise investigation of damaged buildings is essential for effective disaster response. Accordingly, it is expected that the results obtained through this study can be effectively used for the rapid response in disaster management.