• 한국멀티미디어학회논문지
• /
• 제25권1호
• /
• pp.39-51
• /
• 2022

Image-based gender classification and age estimation of human are classic problems in computer vision. Most of researches in this field focus just only one task of either gender classification or age estimation and most of the reported methods for each task focus on accuracy performance and are not computationally light. Thus, running both tasks together simultaneously on low cost mobile or embedded systems with limited cpu processing speed and memory capacity are practically prohibited. In this paper, we propose a novel light-weight gender classification and age estimation method based on ensemble multitasking deep learning with light-weight processing neural network architecture, which processes both gender classification and age estimation simultaneously and in real-time even for embedded systems. Through experiments over various well-known datasets, it is shown that the proposed method performs comparably to the state-of-the-art gender classification and/or age estimation methods with respect to accuracy and runs fast enough (average 14fps) on a Jestson Nano embedded board.

• Smart Structures and Systems
• /
• 제29권1호
• /
• pp.53-62
• /
• 2022

The deployment of advanced structural health monitoring (SHM) systems in large-scale civil structures collects large amounts of data. Note that these data may contain multiple types of anomalies (e.g., missing, minor, outlier, etc.) caused by harsh environment, sensor faults, transfer omission and other factors. These anomalies seriously affect the evaluation of structural performance. Therefore, the effective analysis and mining of SHM data is an extremely important task. Inspired by the deep learning paradigm, this study develops a novel generative adversarial network (GAN) and convolutional neural network (CNN)-based data anomaly detection approach for SHM. The framework of the proposed approach includes three modules : (a) A three-channel input is established based on fast Fourier transform (FFT) and Gramian angular field (GAF) method; (b) A GANomaly is introduced and trained to extract features from normal samples alone for class-imbalanced problems; (c) Based on the output of GANomaly, a CNN is employed to distinguish the types of anomalies. In addition, a dataset-oriented method (i.e., multistage sampling) is adopted to obtain the optimal sampling ratios between all different samples. The proposed approach is tested with acceleration data from an SHM system of a long-span bridge. The results show that the proposed approach has a higher accuracy in detecting the multi-pattern anomalies of SHM data.

• 제어로봇시스템학회논문지
• /
• 제12권2호
• /
• pp.154-160
• /
• 2006

Instantaneous reaction and intelligence are required for autonomous mobile robots to achieve multiple goals in the unpredictable and dynamic environments. Design of the appropriate control architecture and clear definitions of systems are needed to construct and control these robots. This research proposes the hybrid deliberative/reactive control architecture which consists of three layers and uses the method of software structure design. The highest layer, Deliberative Layer makes the overall run-time schedule for navigation and/or manipulation, and the middle layer, Task Execution Layer carries out various missions. The lowest layer, Reactive Layer enables a robot to react rapidly in the dynamic environment and controls the mechanical devices concurrently. This paper proposes independent system supervisors called Manager to reuse the modules so that the Manager supports common use of the system and multi-processing tasks. It is shown that the mobile robot based on the proposed control scheme can perform the basic navigation and cope with the dynamic obstacles reasonably well.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제13권2호
• /
• pp.751-770
• /
• 2019

Action recognition is an essential task in computer vision due to the variety of prospective applications, such as security surveillance, machine learning, and human-computer interaction. The availability of more video data than ever before and the lofty performance of deep convolutional neural networks also make it essential for action recognition in video. Unfortunately, limited crafted video features and the scarcity of benchmark datasets make it challenging to address the multi-person action recognition task in video data. In this work, we propose a deep convolutional neural network-based Effective Hybrid Learning (EHL) framework for two-person interaction classification in video data. Our approach exploits a pre-trained network model (the VGG16 from the University of Oxford Visual Geometry Group) and extends the Faster R-CNN (region-based convolutional neural network a state-of-the-art detector for image classification). We broaden a semi-supervised learning method combined with an active learning method to improve overall performance. Numerous types of two-person interactions exist in the real world, which makes this a challenging task. In our experiment, we consider a limited number of actions, such as hugging, fighting, linking arms, talking, and kidnapping in two environment such simple and complex. We show that our trained model with an active semi-supervised learning architecture gradually improves the performance. In a simple environment using an Intelligent Technology Laboratory (ITLab) dataset from Inha University, performance increased to 95.6% accuracy, and in a complex environment, performance reached 81% accuracy. Our method reduces data-labeling time, compared to supervised learning methods, for the ITLab dataset. We also conduct extensive experiment on Human Action Recognition benchmarks such as UT-Interaction dataset, HMDB51 dataset and obtain better performance than state-of-the-art approaches.

• 한국경영과학회:학술대회논문집
• /
• 한국경영과학회 2001년도 추계학술대회 논문집
• /
• pp.322-325
• /
• 2001

This paper is focused on the development of the DFX system which can support the task that designer or project teams analyse and evaluate the feasibility of the product concept quickly, and then improve the problem of the product and process design before the manufacture in the M-T-O (make-to-order) environments. In this paper, we importantly treated three points for the effective development of the DFX system. The first, we proposed the procedure and architecture of the DFX system which fit with the M-T-O environments. The second, we designed and implemented the DFX prototype system on the WWW, in order to allow the multi-accessing and multi-processing regardless of time and position. The last, we described the benefits of the Web-based DFX system.

• 한국통신학회논문지
• /
• 제28권12B호
• /
• pp.1081-1090
• /
• 2003

차세대 인터넷은 QoS(Quality of Service)의 보장과 초고속 전송을 특징으로 하는데 그 대표적인 구조가 DiffServ(Differentiated Service)와 MPLS(Multi-Protocol Label Switching)이다. 차세대 인터넷에서 종단간 QoS를 보장하기 위해서 DiffServ와 MPLS를 연동하기 위한 기술에 관한 연구가 필수적이고, 현재 이에 관한 연구가 활발하게 진행되고 있다. IETF(Internet Engineering Task Force)의 MPLS WG(Working Group)에서는 이를 위하여 E-LSP(EXP inferred-PSC LSPs)와 L-LSP(Label-Only-Inferred-PSC LSPs) 방안을 제안했는데, 두 가지 방법 모두 MPLS에서 DiffServ의 서비스 클래스를 지원하는데 문제점을 가지고 있다. 본 논문에서는 기존의 E-LPS를 확장하여 DiffServ의 이론적인 모든 클래스와 ECN(Explicit Congestion Notification)을 비롯한 실험적인 기능을 지원하는 확장 E-LSP를 제안한다. 제안된 확장 E-LSP를 이용한 MPLS망에서 보다 세분화된 DiffServ 클래스를 지원할 경우 서비스 품질을 시뮬레이션을 통하여 검증하였다.

• 정보처리학회논문지:소프트웨어 및 데이터공학
• /
• 제12권9호
• /
• pp.399-406
• /
• 2023

클라우드 컴퓨팅은 에지 서버가 동작하는 포그(fog) 레이어가 결합된 에지(edge) 컴퓨팅 아키텍처로 진화하고 있다. 에지 컴퓨팅 아키텍처가 관심을 받는 이유는 짧은 통신 지연으로 실시간 IoT 응용을 지원할 수 있기 때문이다. 이와 동시에 인공지능 기술을 도입한 많은 클라우드 작업 스케줄링 기법들이 제안되었다. 인공지능 기반의 클라우드 작업 스케줄링 기법은 기존 기법보다 더 좋은 성능을 보이지만 스케줄링 시간이 다소 소요된다는 단점이 있다. 이 논문에서는 에지 컴퓨팅 환경에서 분산 딥러닝 학습 기반의 동적 스케줄링 기법을 제안한다. 제안하는 기법은 기존 기법보다 스케줄링 시간이 짧은 장점이 있다. 또한 멀티 에이전트를 통한 분산 딥러닝 학습의 효과성을 보이기 위해 확장적인 실험 환경에서 제안 기법과 기존 인공지능 기법의 성능일 비교 평가하였다. 성능 실험 결과 기존 인공지능 기반 클라우드 작업 스케줄링 기법보다 짧은 스케줄링 시간을 보여 IoT 실시간 응용에 적합함을 보였으며, 확장적인 실험에서도 제안 기법이 완료된 작업의 수에 대하여 우수한 성능을 보임을 증명하였다.

• 제어로봇시스템학회논문지
• /
• 제14권7호
• /
• pp.720-725
• /
• 2008

This paper proposes a system architecture for USN with a service robot to provide more active assisted living services for elderly persons by monitoring their mental and physical well-being with USN environments at home, hospital, or silver town. Sensors embedded in USN are used to detect preventive measures for chronic disease. Logged data are transferred to main controller of a service robot via wireless channel in which the analysis of data is performed. For the purpose of handling emergency situations, it needs real-time processing on gathering variety sensor data, routing algorithms for sensor nodes to a moving sink node and processing of logged data. This paper realized multi-hop sensor network to detect user movements with biometric data transmission and performed algorithms on Xenomai, a real-time embedded Linux. To leverage active sensing, a mobile robot is used of which task was implemented with a priority to process urgent data came from the sink-node. This software architecture is anticipated to integrate sensing, communication and computing with real-time manner. In order to verify the usefulness of a proposed system, the performance of data transferring and processing on a real-time OS with non real-time OS is also evaluated.

• International Journal of Fuzzy Logic and Intelligent Systems
• /
• 제9권4호
• /
• pp.249-254
• /
• 2009

The today's decision making tasks in globalized business and manufacturing become more complex, and ill-defined, and typically multiaspect or multi-discipline due to many influencing factors. The requirement of obtaining fast and reliable decision solutions further complicates the task. Intelligent decision support system (DSS) currently exhibit wide spread applications in business and manufacturing because of its ability to treat ill-structuredness and vagueness associated with complex decision making problems. For multi-dimensional decision problems, generally an optimum single DSS can be developed. However, with an increasing number of influencing dimensions, increasing number of their factors and relationships, complexity of such a system exponentially grows. As a result, software development and maintenance of an optimum DSS becomes cumbersome and is often practically unfeasible for real situations. This paper presents a technically feasible approximation of an optimum DSS through decreasing its complexity by a modular structure. It consists of multiple DSSs, each of which contains the homogenous knowledge's, decision making tools and possibly expertise's pertaining to a certain decision making dimension. Simple, efficient and practical integration mechanism is introduced for integrating the individual DSSs within the proposed overall DSS architecture.

• 대한조선학회논문집
• /
• 제53권5호
• /
• pp.388-399
• /
• 2016

A motion analysis of an underwater towed cable is a complex task due to its nonlinear nature of the problem. The major source of the nonlinearity of the underwater cable analysis is that the motion of the cable involves large rigid-body motion. This large rigid-body motion makes difficult to use standard displacement-based finite element method. In this paper, the authors apply recently developed nodal position-based finite element method which can deal with the geometric nonlinearity due to the large rigid-body motion. In order to enhance the stability of the large-scale nonlinear cable motion simulation, an efficient time-integration scheme is proposed, namely predictor/multi-corrector Newmark scheme. Three different predictors are introduced, and the best predictor in terms of stability and robustness for impulsive cable motion analysis is proposed. As a result, the nonlinear motion of underwater cable is predicted in a very efficient manner compared to the classical finite element of finite difference methods. The efficacy of the method is demonstrated with several test cases, involving static and dynamic motion of a single cable element, and also under water towed cable composed of multiple cable elements.