• Journal of Information Processing Systems
• /
• v.20 no.3
• /
• pp.375-390
• /
• 2024

Edge computing architecture has effectively alleviated the computing pressure on cloud platforms, reduced network bandwidth consumption, and improved the quality of service for user experience; however, it has also introduced new security issues. Existing anomaly detection methods in big data scenarios with cloud-edge computing collaboration face several challenges, such as sample imbalance, difficulty in dealing with complex network traffic attacks, and difficulty in effectively training large-scale data or overly complex deep-learning network models. A lightweight deep-learning model was proposed to address these challenges. First, normalization on the user side was used to preprocess the traffic data. On the edge side, a trained Wasserstein generative adversarial network (WGAN) was used to supplement the data samples, which effectively alleviates the imbalance issue of a few types of samples while occupying a small amount of edge-computing resources. Finally, a trained lightweight deep learning network model is deployed on the edge side, and the preprocessed and expanded local data are used to fine-tune the trained model. This ensures that the data of each edge node are more consistent with the local characteristics, effectively improving the system's detection ability. In the designed lightweight deep learning network model, two sets of convolutional pooling layers of convolutional neural networks (CNN) were used to extract spatial features. The bidirectional long short-term memory network (BiLSTM) was used to collect time sequence features, and the weight of traffic features was adjusted through the attention mechanism, improving the model's ability to identify abnormal traffic features. The proposed model was experimentally demonstrated using the NSL-KDD, UNSW-NB15, and CIC-ISD2018 datasets. The accuracies of the proposed model on the three datasets were as high as 0.974, 0.925, and 0.953, respectively, showing superior accuracy to other comparative models. The proposed lightweight deep learning network model has good application prospects for anomaly traffic detection in cloud-edge collaborative computing architectures.

• Journal of Korean Public Health Nursing
• /
• v.21 no.1
• /
• pp.15-24
• /
• 2007

Purpose: The study aim was to provide basic data needed for formulating systematic visiting nursing strategies by comprehending the characteristics and self-care ability of the object families of public health centers in Korea. Method: The research examined 252 families and 339 family members of the vulnerable class that were registered in a visiting nursing program of an urban public health center. The data of 220 families were analyzed using descriptive analysis, t-test, and ANOVA, after excluding any incomplete data. Result: 1. The most frequent characteristics of families were solitary families (52.8%) and financially vulnerable families (87.3%). The most frequent way of family detection was request of the community office. 2. The most frequent type of family problems were vulnerable families (93.2%), followed by patient families (91.0%). 3. The mean score was 11.67 for family self-care ability. 4. The variables of the number of family members, disease type of the patient family members, and the type of vulnerable family showed a significant difference of family self-care ability. Conclusion: This study suggests that vulnerable families demand specific nursing interventions focused on their own problems and that visiting nurses need to obtain and use supportive resources.

• Journal of Korean Academy of Oral and Maxillofacial Radiology
• /
• v.24 no.2
• /
• pp.439-450
• /
• 1994

Recently, the digital image was introduced into radiological image. The digital image has the power of contrast enhancement, histogram control, and other digitally enhancement. At the point of the resolution, periapical radiograph is superior to the digital image, but enhanced digital procedure improves the diagnostic ability of the digital image. The purpose of this study was to evaluate the diagnostic ability of artificial proximal caries in conventional radiographs, digital radiographs and enhanced digital radiographs (histogram specification). ROC (Receiver Operating Characteristic) analysis and paired t-test were used for the evaluation of detectability, and following results were acquired: 1. The mean ROC area of conventional radiographs was 0.9274. 2. The mean ROC area of unenhanced digital image was 0.9168. 3. The mean ROC area of enhanced digital image was 0.9339. 4. The diagnostic ability of three imaging methods was not significant difference(p>0.05). So, the digital images had similar diagnostic ability of artificial proximal caries to conventional radiographs. If properly enhanced digital image, it may be superior to conventional radiographs.

• Journal of Institute of Control, Robotics and Systems
• /
• v.17 no.6
• /
• pp.521-525
• /
• 2011

Lane marking detection is one of important issues in the field of autonomous mobile robot. Especially, in urban environment, like pavement roads of downtown or tour tracks of Science Park, which have continuous patterns on the surface of the road, the lane marking detection becomes more important ability. Although there were many researches about lane detection and lane tracing, many of them used vision sensors mainly to detect lane marking. In this paper, we obtain 2 dimensional library data of 'Intensity' and 'Distance' using one laser rangefinder only. We design a simple classifier and filtering algorithm for the lane detection which uses only one LRF (Laser Range Finder). Allowing extended usage of LRF, this research provides more functionality not only in range finding but also in lane detecting to mobile robots. This work will be technically helpful for robot developers to design more simple and efficient autonomous driving system using LRF.

• Annual Conference of KIPS
• /
• 2016.10a
• /
• pp.506-509
• /
• 2016

We present real-time pedestrian detection that exploit accuracy of Faster R-CNN network. Faster R-CNN has shown to success at PASCAL VOC multi-object detection tasks, and their ability to operate on raw pixel input without the need to design special features is very engaging. Therefore, in this work we apply and adjust Faster R-CNN to single object detection, which is pedestrian detection. The drawback of Faster R-CNN is its failure when object size is small. Previously, small sized object problem was solved by Scale-aware Network. We incorporate Scale-aware Network to Faster R-CNN. This made our method Scale-aware Faster R-CNN (DF R-CNN) that is both fast and very accurate. We separated Faster R-CNN networks into two sub-network, that is one for large-size objects and another one for small-size objects. The resulting approach achieves a 28.3% average miss rate on the Caltech Pedestrian detection benchmark, which is competitive with the other best reported results.

• Smart Structures and Systems
• /
• v.30 no.3
• /
• pp.303-315
• /
• 2022

The degradation of wheel tread may result in serious hazards in the railway operation system. Therefore, timely wheel defect diagnosis of in-service trains to avoid tragic events is of particular importance. The focus of this study is to develop a novel wheel defect detection approach based on the relevance vector machine (RVM) which enables online detection of potentially defective wheels with trackside monitoring data acquired under different running-speed conditions. With the dynamic strain responses collected by a trackside monitoring system, the cumulative Fourier amplitudes (CFA) characterizing the effect of individual wheels are extracted to formulate multiple probabilistic regression models (MPRMs) in terms of multi-kernel RVM, which accommodate both variables of vibration frequency and running speed. Compared with the general single-kernel RVM-based model, the proposed multi-kernel MPRM approach bears better local and global representation ability and generalization performance, which are prerequisite for reliable wheel defect detection by means of data acquired under different running-speed conditions. After formulating the MPRMs, we adopt a Bayesian null hypothesis indicator for wheel defect identification and quantification, and the proposed method is demonstrated by utilizing real-world monitoring data acquired by an FBG-based trackside monitoring system deployed on a high-speed trial railway. The results testify the validity of the proposed method for wheel defect detection under different running-speed conditions.

• ETRI Journal
• /
• v.44 no.2
• /
• pp.300-312
• /
• 2022

Recently, public safety services have attracted significant attention for their ability to protect people from crimes. Rapid detection of dangerous situations (that is, abnormal situations where someone may be harmed or killed) is required in public safety services to reduce the time required to respond to such situations. This study proposes a novel danger detection technology based on multimodal data, which includes data from multiple sensors (for example, accelerometer, gyroscope, heart rate, air pressure, and global positioning system sensors), and multilog data, which includes contextual logs of humans and places (for example, contextual logs of human activities and crime-ridden districts) over time. To recognize human activity (for example, walk, sit, and punch), the proposed technology uses multimodal data analysis with an attitude heading reference system and long short-term memory. The proposed technology also includes multilog data analysis for detecting whether recognized activities of humans are dangerous. The proposed danger detection technology will benefit public safety services by improving danger detection capabilities.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2022.05a
• /
• pp.154-157
• /
• 2022

In the existing Tensorflow CNN environment, the object detection method is a method of performing object labeling and detection by Tensorflow itself. However, with the advent of YOLO, the efficiency of image object detection has increased. As a result, more deep layers can be built than existing neural networks, and the image object recognition rate can be increased. Therefore, in this paper, the detection ability and speed were compared and analyzed by designing an object detection system based on Darknet and YOLO and performing multi-layer construction and learning based on the existing convolutional neural network. For this reason, in this paper, a neural network methodology that efficiently uses Darknet's learning is presented.

• The Journal of Pediatrics of Korean Medicine
• /
• v.15 no.1
• /
• pp.77-104
• /
• 2001

The effects of the oriental herbal medicine Saenghyetang(SHT, 生慧湯), which consists of Rehmanniae Radix (熟地黃 九蒸: was made by 9th steam) 40g, Corni Fructus(山茱黃) 16g, Polygalae Radix(遠志) 8g, Zizyphi Spinosae Semen(酸棗仁) 2g, Biotae Semen(柏子仁 去油: oil ingredient was removed) 20g, Poria Cocos(茯笭) 12g, Ginseng Radix(人蔘) 12g, Acori Graminei Rhizoma(石菖蒲) 2g, Sinapis Semen(白芥子) 8g, on learning ability and memory were investigated. Hot water extract(HWE) and ethanol extract(EE) from SHT were used for the studies. Learning ability and memory are related to modifications of synaptic strength among neurons that interactive. Enhanced synaptic coincidence detection leads to improved learning ability and memory. If the NMDA receptor, a synaptic coincidence detector, acts as a graded switch for memory formations, enhanced signal detection by NMDA receptors should enhance learning ability and memory. It was shown that NR2B was increased in the forebrains of oriental medicine-administrated mice, leading to enhanced activation of NMDA receptors and facilitating synaptic potentiation in response to stimulation at 10-100 Hz. These HWE-SHT treated mice exhibited that superior ability in learning and memory when performing various behavioral tasks, showing that NR2B is enhanced by HWE-SHT treatment and also is critical in gating the age-dependent threshold for plasticity and memory formation. NMDA receptor-dependent modifications, which were mediated in part by HWE administration, of synaptic efficacy, therefore, represent a mechanism for associative learning ability and memory. Results suggest that oriental medical enhancement of NR2B contributes to increase intelligence and memory in mammals On the other hand, to examine the effects of EE-SHT on the learning ability and memory in experimental mice, EE-SHT was tested on passive and active avoidance responses. The EE-SHT ameliorated the memory retrieval deficit induced by ethanol in mice, but not other memory impairments. EE-SHT(10, 20mg/100 g, p.o.) did not affect the passive avoidance responses of normal mice in the step through and step down tests, the conditioned and unconditioned avoidance responses of normal mice in the shuttle box, lever press performance tests and the ambulatory activity of normal mice in a normal condition. However, EE-SHT at 20 mg/kg significantly decrease the spontaneous motor activity during the shuttle box test, and also to extend the sleeping time induced by pentobarbital in mice. These results suggest that SHT has an ameliorating effect on memory retrieval impairments and a weak tranquilizing action.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.33 no.11A
• /
• pp.1063-1071
• /
• 2008

This paper proposes a prior maximum likelihood (ML) detection verifier which has an ability to verify if the zero forcing (ZF) detection results are identical to the ML detection results. Since more than 90% of ZF detection results are identical to ML detection results, the proposed verifier makes it possible to omit the computationally complex ML detection in 90% cases of MIMO signal detections. The proposed verifier is designed by using the diversity gain obtained from converting MIMO signal into single input multiple output (SIMO) signals. In the proposed method, single input multiple output (SIMO) signals for each transmit antenna are separated from MIMO signals after the MIMO signals are detected by ZF method. Computer simulations show that the true alarm probability of the proposed verifier is more than 80% and the false alarm probability is less than $10^{-4}$.