• KIPS Transactions on Software and Data Engineering
• /
• v.10 no.11
• /
• pp.491-500
• /
• 2021

In this study, we propose a method for detecting fall behavior using MS Kinect v2 RGBD Camera-based Human-Skeleton Keypoints and a 2-Stacked Bi-LSTM model. In previous studies, skeletal information was extracted from RGB images using a deep learning model such as OpenPose, and then recognition was performed using a recurrent neural network model such as LSTM and GRU. The proposed method receives skeletal information directly from the camera, extracts 2 time-series features of acceleration and distance, and then recognizes the fall behavior using the 2-Stacked Bi-LSTM model. The central joint was obtained for the major skeletons such as the shoulder, spine, and pelvis, and the movement acceleration and distance from the floor were proposed as features of the central joint. The extracted features were compared with models such as Stacked LSTM and Bi-LSTM, and improved detection performance compared to existing studies such as GRU and LSTM was demonstrated through experiments.

• Journal of IKEEE
• /
• v.25 no.4
• /
• pp.762-765
• /
• 2021

In this paper, an automotive ECU (electronic control unit) with Kalman filter accelerator is designed and implemented. RISC-V is exploited as a processor core. Accelerator for Kalman filter matrix operation, CAN (controller area network) controller for in-vehicle network, and LIN (local interconnect network) controller are designed and embedded. Kalman filter operation consists of time update process and measurement update process. Current state variable and its error covariance are estimated in time update process. Final values are corrected from input measurement data and Kalman gain in measurement update process. Usually floating-point multiplication is exploited in software implementation, but fixed-point multiplier considering accuracy analysis is exploited in this paper to reduce hardware area. In 28nm silicon fabrication, its operating frequency, area, and gate counts are 100MHz, 0.37mm², and 760k gates, respectively.

• Imaging Science in Dentistry
• /
• v.51 no.4
• /
• pp.429-438
• /
• 2021

Purpose: This study investigated the accuracy of laser-scanned models and 3-dimensional(3D) rendered cone-beam computed tomography (CBCT) compared to the gold standard (plaster casts) for linear measurements on dental arches. Materials and Methods: CBCT scans and plaster models from 30 patients were retrieved. Plaster models were scanned by an Emerald laser scanner (Planmeca, Helsinki, Finland). Sixteen different measurements, encompassing the mesiodistal width of teeth and both arches' length and width, were calculated using various landmarks. Linear measurements were made on laser-scanned models using Autodesk Meshmixer software v. 3.0 (Autodesk, Mill Valley, CA, USA), on 3D-rendered CBCT models using OnDemand 3D v. 1.0 (Cybermed, Seoul, Korea) and on plaster casts by a digital caliper. Descriptive statistics, the paired t-test, and intra- and inter-class correlation coefficients were used to analyze the data. Results: There were statistically significant differences between some measurements on plaster casts and laser-scanned or 3D-rendered CBCT models (P<0.05). Molar mesiodistal width and mandibular anterior arch width deviated significantly different from the gold standard in both methods. The largest mean differences of laser-scanned and 3D-rendered CBCT models compared to the gold standard were 0.12±0.23 mm and 0.42±0.53 mm, respectively. Most of the mean differences were not clinically significant. The intra- and inter-class correlation results were acceptable for all measurements(>0.830) and between observers(>0.801). Conclusion: The 3D-rendered CBCT images and laser-scanned models were useful and accurate alternatives to conventional plaster models. They could be used for clinical purposes in orthodontics and prostheses.

• Journal of IKEEE
• /
• v.26 no.3
• /
• pp.468-474
• /
• 2022

In this paper, we propose a method to implement a real-time fire alarm system using deep learning. The deep learning image dataset for fire alarms acquired 1,500 sheets through the Internet. If various images acquired in a daily environment are learned as they are, there is a disadvantage that the learning accuracy is not high. In this paper, we propose a fire image data expansion method to improve learning accuracy. The data augmentation method learned a total of 2,100 sheets by adding 600 pieces of learning data using brightness control, blurring, and flame photo synthesis. The expanded data using the flame image synthesis method had a great influence on the accuracy improvement. A real-time fire detection system is a system that detects fires by applying deep learning to image data and transmits notifications to users. An app was developed to detect fires by analyzing images in real time using a model custom-learned from the YOLO V4 TINY model suitable for the Edge AI system and to inform users of the results. Approximately 10% accuracy improvement can be obtained compared to conventional methods when using the proposed data.

• Geomechanics and Engineering
• /
• v.30 no.1
• /
• pp.67-73
• /
• 2022

In this study, the fiber orientation distribution (FOD) is investigated using both micro-CT (computerized tomography) and image analysis of physically cut specimens prepared from Polyvinyl Alcohol (PVA) fiber reinforced cemented Toyoura sand. The micro-CT images of the fiber reinforced cemented sand specimens were visualized in horizontal and vertical sections. Scans were obtained using a frame rate of two frames and an exposure time of 500 milliseconds. The number of images was set to optimize and typically resulted in approximately 3000 images. Then, the angles of the fibers for horizontal sections and in vertical section were calculated using the VGStudio MAX software. The number of fibers intersecting horizontal and vertical sections are counted using these images. A similar approach was used for physically cut specimens. The variation of results of fiber orientation between micro-CT scans and visual count were approximately 4-8%. The micro-CT scans were able to precisely investigate the fiber orientation distribution of fibers in these samples. The results show that 85-90% of the PVA fibers are oriented between ±30° of horizontal, and approximately 95% of fibers have an orientation that lies within ±45° of the horizontal plane. Finally, a comparison of experimental results with the generalized fiber orientation distribution function 𝜌(θ) is presented for isotropic and anisotropic distribution in fiber reinforced cemented Toyoura sand specimens. Experimentally, it can be seen that the average ratio of the number of fibers intersecting the finite area on a vertical plane to number of fibers intersecting the finite area on a horizontal plane (N^V_tot/N^H_tot) cut through a sample varies from 2.08 to 2.12 (an average ratio of 2.10 is obtained in this study). Based up on the analytical predictions, it can be seen that the average N^V_tot/N^H_tot ratio varies from 2.13 to 2.17 for varying n values (an average ratio of 2.15).

• Advances in Computational Design
• /
• v.7 no.1
• /
• pp.57-68
• /
• 2022

A simulation study of the distribution of magnetic flux induced by a U-shaped electromagnet into a two-layer massive object with variations in the depth and properties of the surface layer has been carried out. It has been established that the hardened surface layer "pushes" the magnetic flux into the bulk of the magnetized object and the magnetic flux penetration depth monotonically increases with increasing thickness of the hardened layer. A change in the thickness and magnetic properties of the surface layer leads to a redistribution of magnetic fluxes passing between the poles of the electromagnet along with the layer and the bulk of the steel object. In this case, the change in the layer thickness significantly affects the magnitude of the tangential component of the field on the surface of the object in the interpolar space, and the change in the properties of the layer affects the magnitude of the magnetic flux in the magnetic "transducer-object" circuit. This difference in magnetic parameters can be used for selective testing of the surface hardening quality. It has been shown that the hardened layer pushes the magnetic flux into the depth of the magnetized object. The nominal depth of penetration of the flow monotonically increases with an increase in the thickness of the hardened layer.

• KNOM Review
• /
• v.24 no.1
• /
• pp.1-12
• /
• 2021

VM (Virtual Machine) live migration is a server virtualization technique for deploying a running VM to another server node while minimizing downtime of a service the VM provides. Currently, in cloud data centers, VM live migration is widely used to apply load balancing on CPU workload and network traffic, to reduce electricity consumption by consolidating active VMs into specific location groups of servers, and to provide uninterrupted service during the maintenance of hardware and software update on servers. It is critical to use VMlive migration as a prevention or mitigation measure for possible failure when its indications are detected or predicted. In this paper, we propose two VNF live migration methods; one for predictive load balancing and the other for a proactive measure in failure. Both need machine learning models that learn periodic monitoring data of resource usage and logs from servers and VMs/VNFs. We apply the second method to a vEPC (Virtual Evolved Pakcet Core) failure scenario to provide a detailed case study.

• Journal of the Korean Geotechnical Society
• /
• v.23 no.4
• /
• pp.133-147
• /
• 2007

Downhole seismic method is very economic and easy of operation because it uses only one borehole and simple surface source to obtain the shear wave velocity ($V_s$) profile of a site. Even though it is widely used by the site investigation companies, universities and institutes, however, the $V_s$ profile determined by downhole seismic method has often low reliability due to employment of wrong combinations of field losing equipment and interpretation method and deficiency of experience. Round robin test was performed and testing equipment and procedure were compared. Adequate downhole seismic testing equipment was constructed based on the comparison and verification study of the round robin test. The data acquisition and software interpretation were also developed for automation and quick test in field. Finally, the effectiveness and applicability were verified through the field test by using the constructed testing system.

• Animal Bioscience
• /
• v.36 no.6
• /
• pp.980-989
• /
• 2023

Objective: Iris pattern recognition system is well developed and practiced in human, however, there is a scarcity of information on application of iris recognition system in animals at the field conditions where the major challenge is to capture a high-quality iris image from a constantly moving non-cooperative animal even when restrained properly. The aim of the study was to validate and identify Black Bengal goat biometrically to improve animal management in its traceability system. Methods: Forty-nine healthy, disease free, 3 months±6 days old female Black Bengal goats were randomly selected at the farmer's field. Eye images were captured from the left eye of an individual goat at 3, 6, 9, and 12 months of age using a specialized camera made for human iris scanning. iGoat software was used for matching the same individual goats at 3, 6, 9, and 12 months of ages. Resnet152V2 deep learning algorithm was further applied on same image sets to predict matching percentages using only captured eye images without extracting their iris features. Results: The matching threshold computed within and between goats was 55%. The accuracies of template matching of goats at 3, 6, 9, and 12 months of ages were recorded as 81.63%, 90.24%, 44.44%, and 16.66%, respectively. As the accuracies of matching the goats at 9 and 12 months of ages were low and below the minimum threshold matching percentage, this process of iris pattern matching was not acceptable. The validation accuracies of resnet152V2 deep learning model were found 82.49%, 92.68%, 77.17%, and 87.76% for identification of goat at 3, 6, 9, and 12 months of ages, respectively after training the model. Conclusion: This study strongly supported that deep learning method using eye images could be used as a signature for biometric identification of an individual goat.

• Journal of the Korean Institute of Electrical and Electronic Material Engineers
• /
• v.36 no.2
• /
• pp.158-163
• /
• 2023

Recently, the global demands for high voltage power semiconductors are increasing across various industrial fields. The use of electric cars with high safety and convenience is becoming practical, and IGBT modules of 3.3 kV and 1.2 kA or higher are used for electric locomotives. Delicate design and advanced process technology are required, and research on the optimization of high-voltage IGBT parts is urgently needed in the industry. In this study, we attempted to design a simulation process through TCAD (technology computer-aid design) software to optimize the process conditions of the fielding process among the core unit processes for an especial high yield voltage. As well, the prior circuit technology design and a ring pattern with a large number of ring formation structures outside the wafer similar to the chip structure of other companies were constructed for 3.3 kV NPT-IGBT through a unit process demonstration experiment. The ring pattern was designed with 21 rings and the width of the ring was 6.6 ㎛. By changing the spacing between patterns from 17.4 ㎛ to 35.4 ㎛, it was possible to optimize the spacing from 19.2 ㎛ to 18.4 ㎛.