• International Journal of Advanced Culture Technology
• /
• 제10권2호
• /
• pp.246-251
• /
• 2022

As the number of data and devices explodes, centralized data processing and AI analysis have limitations due to the load on the network and cloud. On-device AI technology can provide intelligent services without overloading the network and cloud because the device itself performs AI models. Accordingly, the need for on-device AI technology is emerging. Many smartphones are equipped with On-Device AI technology to support the use of related functions. In this paper, we compare software tools that implement On-Device AI.

• 한국멀티미디어학회논문지
• /
• 제15권12호
• /
• pp.1485-1491
• /
• 2012

In this paper, we address the problem of fast handover support for future internet networks. The current mobility support protocols, for example Proxy Mobile IPv6, follow a centralized architecture in which a mobility controller is used to manage intra-network handovers that gives rise to high handover latency. To handle this problem, we propose a novel distributed mobility control architecture in which gateways in the network can exchange user information with neighbour gateways to support intra-network handovers, the mobility controller is only needed in case of inter-network handovers. Simulation results show that our architecture reduce approximately 20% intra-network handover latency than Proxy Mobile IPv6.

• 스마트미디어저널
• /
• 제5권4호
• /
• pp.41-48
• /
• 2016

In mobile ad hoc networks(MANETs), all the nodes in a network have limited resources. Therefore, communication topology which has long lifetime is suitable for nodes in MANETs. And MANETs are exposed to various threats because of a new node which can join the network at any time. There are various researches on security problems in MANETs and many researches have tried to make efficient schemes for reducing network power consumption. Power consumption is necessary to secure networks, however too much power consumption can be critical to network lifetime. This paper focuses on energy efficient monitoring node distribution for enhancing network lifetime in MANETs. Since MANETs cannot use centralized infrastructure such as security systems of wired networks, we propose an efficient IDS node distribution scheme using minimum spanning tree (MST) method to cover all the nodes in a network and enhance the network lifetime. Simulation results show that the proposed algorithm has better performance in comparison with the existing algorithms.

• ETRI Journal
• /
• 제26권1호
• /
• pp.27-37
• /
• 2004

The architecture in a differentiated services (DiffServ) network is based on a simple model that applies a per-class service in the core node of the network. However, because the network behavior is simple, the network structure and provisioning is complicated. If a service provider wants dynamic provisioning or a better bandwidth guarantee, the differentiated services network must use a signaling protocol with QoS parameters or an admission control method. Unfortunately, these methods increase the complexity. To overcome the problems with complexity, we investigated scalable dynamic provisioning for admission control in DiffServ networks. We propose a new scalable $qDPM^2$ mechanism based on a centralized bandwidth broker and distributed measurement-based admission control and movable boundary bandwidth management to support heterogeneous QoS requirements in DiffServ networks.

• 한국정보처리학회:학술대회논문집
• /
• 한국정보처리학회 2004년도 춘계학술발표대회
• /
• pp.1529-1532
• /
• 2004

A Mobile Ad Hoc NETwork (MANET) is a collection of wireless mobile nodes that forms a temporary network without the need for any existing network infrastructure or centralized administration. Therefore, such a network is designed to operate in a highly dynamic environment due to node mobility. In mobile ad hoc network, frequent topological changes cause routing a challenging problem and without the complete view of the network topology, establishing the shortest path from the source node to the destination node is difficult. In this paper, we suggest a routing approach which utilizes location information to setup the shortest possible path between the source node and the destination node. Location information is obtained through Global Positioning System (GPS) and this geographical coordinate information of the destination node is used by the source node and intermediate nodes receiving route request messages to determine the shortest path to the destination from current node.

• 한국정보통신학회논문지
• /
• 제9권1호
• /
• pp.165-171
• /
• 2005

본 논문에서는 중앙집중형 동적부하균등을 효율적으로 처리하기 위하여 네트워크기반 병렬 유전자 알고리즘을 이용하였다. 기존의 유전자 알고리즘을 적용한 경우와는 달리 클라이언트들에서 최적작업 할당의 탐색을 분산처리하여 중앙 스케줄러의 성능을 향상시킬 수 있었다. 최적해의 수렴속도를 향상시키기 위해 선택연산은 룰렛휠 선택과 엘리트 보존전략을 함께 사용하였고, 염색체 인코딩은 슬라이딩윈도우기법을 이용하였으며 교차연산은 주기교차방법을 이용하였다. 부하균등기법의 유연성 변화에 따른 중앙 스케줄러의 성능을 모의실험한 결과 기존의 방법보다 성능이 향상됨을 확인하였다.

• 한국정보과학회논문지:정보통신
• /
• 제32권5호
• /
• pp.574-582
• /
• 2005

무선 센서 네트워크에서 노드들의 위치를 측정하는 문제는 사건탐지, 라우팅, 정보추적 등의 중요한 네트워크 기능을 수행하기 위해 필수적으로 해결해야 한다. 위치측정 문제는 노드들 간의 연결성이 알려져 있을 때 네트워크의 모든 노드의 위치를 결정하는 문제이다. 본 논문에서는 분산 알고리즘인 중점기법을 중앙 집중적인 알고리즘으로 확장한다. 확장된 알고리즘은 단순하다는 중점 기법의 장점을 지니면서도 각 미지 노드가 세 개 이상의 고정 노드들의 중첩된 전송 범위에 속해야 한다는 단점을 갖지 않는다. 본 논문의 알고리즘은 위치 측정 문제가 일차원 행렬 방정식으로 정형화 될 수 있음을 보여준다. 이러한 일차원 행렬 방정식이 유일한 해를 가짐을 수학적으로 증명함으로써 모든 미지 노드들의 위치를 유일하게 결정할 수 있음을 보인다.

• 한국정보처리학회:학술대회논문집
• /
• 한국정보처리학회 2015년도 추계학술발표대회
• /
• pp.426-429
• /
• 2015

최근 급격히 증가한 모바일 기기로 인하여 발생되는 데이터/제어 트래픽은 LTE/EPC 네트워크에서 중앙에 과다한 트래픽 수용문제가 중요 이슈로 부각되고 있다. 기존의 Centralized Mobility Management(CMM) 기반의 LTE/EPC 네트워크에서 Mobility Anchor 역할을 수행하는 Packet Data Network Gateway (P-GW)에서는 데이터 트래픽 과부하가 발생한다. 또한 Distributed Mobility Management (DMM) 기반의 LTE/EPC 네트워크에서 분산된 Mobility Anchor 역할을 수행하는 PDN Edge Gateway (P-EGW)에서는 제어 트래픽의 과부하가 발생한다. 본 논문에서는 이러한 문제를 해결하기 위하여 CMM 기반과 DMM 기반을 결합한 새로운 Software Defined Network (SDN) 기반의 LTE/EPC 네트워크 이동성 관리 기법을 제안한다. 이를 위하여, P-EGW를 네트워크 내에 분산 배치하고 중앙에 P-GW를 배치한다. SDN 컨트롤러는 EPC의 역할도 수행하며 UE의 이동성에 따라 적절한 CMM 기법과 DMM 기법을 이용하도록 하는 기법을 제안한다. 또한, 제안하는 새로운 LTE/EPC 네트워크 구조와 기존의 CMM기반의 LTE/EPC 네트워크 구조, DMM 기반의 LTE/EPC 네트워크 구조를 핸드오버 지연시간과 데이터 전송시간 측면에서 성능 비교 분석을 한다.

• 한국정보시스템학회지:정보시스템연구
• /
• 제28권1호
• /
• pp.73-96
• /
• 2019

Purpose The purpose of this study is to examine the effect of communication-structure characteristics on performance in online collaboration using the data from Github, one of representative open source software development platforms. We analyze the impact of in-degree/out-degree centralization and reciprocity of communication network on collaboration performance in each project. In addition, we investigate the moderating effect of owner types, an individual developer or an organization. Design/methodology/approach We collect the data of 838 Github projects, and conduct social network analysis for measuring in-degree/out-degree centralization and reciprocity as independent variables. With these variables, hierarchical regression analysis is employed on the relationship between the characteristics of communication structure and collaborative performance. Findings Our results show that for the project owned by an organization, the centralized structure of communication is not associated with the collaboration performance. In addition, the reciprocity is positively related to the collaboration performance. On the other hand, for the project owned by an individual developer, the centralized structure of communication is positively related to the performance, and the reciprocity does not show the positive relationship on the performance.

• International Journal of Computer Science & Network Security
• /
• 제21권8호
• /
• pp.127-136
• /
• 2021

Real-time human recognition is a challenging task, as the images are captured in an unconstrained environment with different poses, makeups, and styles. This limitation is addressed by generating several facial images with poses, makeup, and styles with a single reference image of a person using Generative Adversarial Networks (GAN). In this paper, we propose deep learning-based human recognition using integration of GAN and Spatial Domain Techniques. A novel concept of human recognition based on face depiction approach by generating several dissimilar face images from single reference face image using Domain Transfer Generative Adversarial Networks (DT-GAN) combined with feature extraction techniques such as Local Binary Pattern (LBP) and Histogram is deliberated. The Euclidean Distance (ED) is used in the matching section for comparison of features to test the performance of the method. A database of millions of people with a single reference face image per person, instead of multiple reference face images, is created and saved on the centralized server, which helps to reduce memory load on the centralized server. It is noticed that the recognition accuracy is 100% for smaller size datasets and a little less accuracy for larger size datasets and also, results are compared with present methods to show the superiority of proposed method.