• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.14 no.11
• /
• pp.4426-4442
• /
• 2020

Since scene segmentation is becoming a hot topic in the field of autonomous driving and medical image analysis, researchers are actively trying new methods to improve segmentation accuracy. At present, the main issues in image semantic segmentation are intra-class inconsistency and inter-class indistinction. From our analysis, the lack of global information as well as macroscopic discrimination on the object are the two main reasons. In this paper, we propose a Densely connected residual Attention network (DA-Res2Net) which consists of a dense residual network and channel attention guidance module to deal with these problems and improve the accuracy of image segmentation. Specifically, in order to make the extracted features equipped with stronger multi-scale characteristics, a densely connected residual network is proposed as a feature extractor. Furthermore, to improve the representativeness of each channel feature, we design a Channel-Attention-Guide module to make the model focusing on the high-level semantic features and low-level location features simultaneously. Experimental results show that the method achieves significant performance on various datasets. Compared to other state-of-the-art methods, the proposed method reaches the mean IOU accuracy of 83.2% on PASCAL VOC 2012 and 79.7% on Cityscapes dataset, respectively.

• ETRI Journal
• /
• v.40 no.5
• /
• pp.582-591
• /
• 2018

In this paper, we present a performance analysis of large-scale multi-input multi-output (MIMO) systems for wireless backhaul networks. We focus on fully connected N nodes in a wireless meshed and multi-hop network topology. We also consider a large number of antennas at both the receiver and transmitter. We investigate the transmission schemes to support fully connected N nodes for half-duplex and full-duplex transmission, analyze the achievable ergodic sum rate among N nodes, and propose a closed-form expression of the achievable ergodic sum rate for each scheme. Furthermore, we present numerical evaluation results and compare the resuts with closed-form expressions.

• ETRI Journal
• /
• v.39 no.2
• /
• pp.275-283
• /
• 2017

A green technology for reducing energy consumption has become a critical factor in ICT industries. However, for the telecommunications sector in particular, most network elements are not usually optimized for power efficiency. Here, we propose a novel energy-efficient packet switching method for use in an IP network for reducing unnecessary energy consumption. As a green networking approach, we first classify the network nodes into either header or member nodes. The member nodes then put the routing-related module at layer 3 to sleep under the assumption that the layer in the OSI model can operate independently. The entire set of network nodes is then partitioned into clusters consisting of one header node and multiple member nodes. Then, only the header node in a cluster conducts IP routing and its member nodes conduct packet switching using a specially designed identifier, a tag. To investigate the impact of the proposed scheme, we conducted a number of simulations using well-known real network topologies and achieved a more energy- efficient performance than that achieved in previous studies.

• Proceedings of the IEEK Conference
• /
• 2002.07c
• /
• pp.2079-2082
• /
• 2002

An interconnection network is called hamiltonian-connected if there exists a hamiltonian path joining every pair of nodes. We consider the problem of adding edges to a mesh to make it hamiltonian- connected. We show that at least two edges are necessary for the problem. Also, we present the method to add two edges to a mesh so that the resulting network is hamiltonian-connected.

• Proceedings of the KIPE Conference
• /
• 2008.06a
• /
• pp.277-279
• /
• 2008

In this paper, a novel active boost converter for SR drive is proposed. An active capacitor circuit is added in the front-end. Based on this active capacitor network, when boost switch turns off, this network seems as passive capacitor network. And the voltage of boost capacitor can keep balance with dc-link voltage automatically. In the capacitor network, discharging voltage is general dc-link voltage in parallel-connected capacitors; charging voltage is double dc-link voltage in series-connected capacitors. When boost switch turns on, two capacitors are connected in series, and discharging voltage is up to double dc-link voltage. So the fast excitation current can be obtained from this mode. Profit from fast excitation and fast demagnetization mode, the performance and output power can be improved. Some computer simulations are done to verify the performance of proposed converter.

• The Journal of the Korea Contents Association
• /
• v.13 no.12
• /
• pp.519-529
• /
• 2013

This paper introduces a moving object tracing system using network-connected CCTV and smartphone. In an emergency situation, a smartphone of the reporter gains its GPS information and sends that to the central server. The central server stores received GPS information as a dangerous area in the database and keeps sending the order of tracing the reporter to a network-connected CCTV which can film the reporter. At the same time, the central server sends pictures of the reporter to his or her family and related organization in order to handle the emergency situation as soon as possible. In addition, when a reporter want to know the risk around destination, the central server informs dangerous areas to the reporter by using smart phone application and database of a danger spot.

• Journal of the Korea Computer Graphics Society
• /
• v.27 no.5
• /
• pp.73-79
• /
• 2021

Recently, in the field of computer animation, a method for generating motion using deep learning has been studied away from conventional finite-state machines or graph-based methods. The expressiveness of the network required for learning motions is more influenced by the diversity of motion contained in it than by the simple length of motion to be learned. This study aims to find an efficient network structure when the types of motions to be learned are diverse. In this paper, we train and compare three types of networks: basic fully-connected structure, mixture of experts structure that uses multiple fully-connected layers in parallel, recurrent neural network which is widely used to deal with seq2seq, and transformer structure used for sequence-type data processing in the natural language processing field.

• Journal of Power Electronics
• /
• v.18 no.3
• /
• pp.746-756
• /
• 2018

Harmonic resonance exists in grid-connected inverter systems. In order to determine the network components that contribute to harmonic resonance and the composition of the resonant circuit, sensitivity theory is applied to the resonance characteristic analysis. Based on the modal analysis, the theory of sensitivity is applied to derive a formula for determining the sensitivities of each network component parameter under a resonance circumstance that reflects the participation of the network component. The solving formula is derived for both parallel harmonic resonance and series harmonic resonance. This formula is adopted to a 4-node grid-connected test system. The analysis results reveal that for a certain frequency, the participation of parallel resonance and series resonance are not the same. Finally, experimental results demonstrate that the solving formula for sensitivity is feasible for grid-connected systems.

• Journal of KIISE
• /
• v.43 no.12
• /
• pp.1420-1427
• /
• 2016

This paper proposes a method taking advantage of Fog computing and SDN in the connected vehicle environment which is having an unstable communication channel and a dynamic topology. For this purpose, the controller should understand the current state of the overall network by maintaining recent network topology, especially, the mobility information of mobile nodes. These are managed by the controller, and are important in unstable conditions in the mobile environment. The mobility levels are divided into 3 categories. We can efficiently exploit that information. By utilizing network state information, we suggest two outcomes. First, we reduce the control message overhead by adjusting the period of beacon messages. Second, we propose a recovery process to prepare the communication failure. We can efficiently recover connection failure through mobility information. Furthermore, we suggest a path recovery by decoupling the cloud level and the fog level in accordance with application data types. The simulation results show that the control message overhead and the connection failure time are decreased by approximately 55% and 5%, respectively in comparison to the existing method.

• Convergence Security Journal
• /
• v.21 no.1
• /
• pp.79-92
• /
• 2021

In a society where ICT technology is converged, the use of various network-connected devices such as IoT is spreading. Network-connected devices are inevitably exposed to the threat of hacking such as information leakage, and countermeasures need to be prepared to respond. Security certification system for IoT devices has been introduced to promote security of IoT products, and for this purpose, legalization and standardization of certification standards and methods are in progress. In line with this, in Korea, as the Information and Communication Network Act was revised in 2020, ICT convergence devices connected to the network were newly defined as "information and communication network connected devices," and the basis for the security certification system is being established. We summarized related domestic and foreign trends and suggest specific considerations for implementing the security certification system for IoT devices in South Korea.