• 반도체디스플레이기술학회지
• /
• 제18권2호
• /
• pp.1-5
• /
• 2019

As the number of IOT devices is growing rapidly, various 'see-thru connection' techniques have been reported for efficient communication with them. In this paper, we propose a deep learning based IOT device recognition system for interaction with these devices. The overall system consists of a TensorFlow based deep learning server and two Android apps for data collection and recognition purposes. As the basic neural network model, we adopted Google's inception-v3, and modified the output stage to classify 20 types of IOT devices. After creating a data set consisting of 1000 images of 20 categories, we trained our deep learning network using a transfer learning technology. As a result of the experiment, we achieve 94.5% top-1 accuracy and 98.1% top-2 accuracy.

• ETRI Journal
• /
• 제41권5호
• /
• pp.560-573
• /
• 2019

Two supervised learning algorithms, a basic neural network and a long short-term memory recurrent neural network, are applied to traffic including DDoS attacks. The joint effects of preprocessing methods and hyperparameters for machine learning on performance are investigated. Values representing attack characteristics are extracted from datasets and preprocessed by two methods. Binary classification and two optimizers are used. Some hyperparameters are obtained exhaustively for fast and accurate detection, while others are fixed with constants to account for performance and data characteristics. An experiment is performed via TensorFlow on three traffic datasets. Three scenarios are considered to investigate the effects of learning former traffic on sequential traffic analysis and the effects of learning one dataset on application to another dataset, and determine whether the algorithms can be used for recent attack traffic. Experimental results show that the used preprocessing methods, neural network architectures and hyperparameters, and the optimizers are appropriate for DDoS attack detection. The obtained results provide a criterion for the detection accuracy of attacks.

• Investigative Magnetic Resonance Imaging
• /
• 제17권2호
• /
• pp.123-132
• /
• 2013

목적 : 정상인 경수에 대한 내부 부위별, 경수 레벨별 확산텐서영상과 뇌척수액 속도에 대한 정상인 기준을 마련하고자 하였다. 대상 및 방법 : 정상인을 대상으로 3T MRI를 이용하여 영상을 획득하였다. 확산텐서영상은 척수 레벨별로 각각 미리 정의해둔 백질의 후삭, 좌, 우 측삭과 회백질에 관심영역을 정하여 FA, MD에 대한 평균값을 구하였다. PC 기법을 이용한 뇌척수액 속도 측정은 C2-3, C4-5, C5-6 척수레벨을 각각 스캔한 후, 척수 레벨별 최대 수축기 및 이완기 속도를 측정하였다. 결과 : FA 값과 MD 값에서 회백질과 백질의 내부 구조 사이에서는 통계적으로 유의한 차이 (p < 0.05)를 보였다. 경수 레벨별에서는 회백질의 FA 값에서 통계적으로 유의한 차이(p < 0.05)를 보였으며, 나머지 부분에서는 통계적으로 유의한 차이를 보이지 않았다 (p > 0.05). 뇌척수액 속도 측정 결과, 평균 최대 수축기 속도 $5.18{\pm}2.00cm/sec$, 평균 최대 이완기 속도 $-7.32{\pm}3.18cm/sec$ 이었으며, 경수 레벨별로는 통계적으로 유의한 차이를 보이지 않았다 (p > 0.05). 결론 : 정상인에 대한 경수 확산텐서영상의 정량적 수치 및 경수 레벨에 영향을 받지 않는 뇌척수액 속도에 대한 평균치를 알 수가 있었다.

• 대한의용생체공학회:의공학회지
• /
• 제25권1호
• /
• pp.43-47
• /
• 2004

뇌의 기능성과 분산 영역간의 연결성에 대한 관심이 매우 증대되고 있는 추세다. 그 중에서 DTI (확산 텐서 영상)는 비침습적으로 뇌에서의 물분자의 확산과 뇌의 회백질 신경 다발 구조에 대한 정보를 제공할 수 있는 자기공명영상 기법이다. 따라서 우리는 이 기법을 통해 다른 해부학적 혹은 기능적 자기공명영상 기법으로는 얼을 수 없는 뇌의 신경 섬유 다발과 대뇌 피질 영역의 연결성 정보를 얻을 수 있다. 본 연구에서는 컬러 코딩 방식에 의하여 뇌의 주된 신경 다발의 방향성을 가시화 하고자 한다. 인간 뇌의 방향성 지도는 연합섬유, 방사섬유. 교련섬유. 운동 및 감각회로 섬유의 다발들을 구별하기 쉽게 하여준다. 우리는 이 목적 하에 24비트 칼라코딩 방법을 윈도우 PC 환경에서 IDL을 이용하여 구현하였다. 덧붙여 방향성의 각각의 성분 및 이방성에 대한 컬러 코딩과 이들을 표현하기 위한 다양한 색상표를 구현하였다. 결론적으로 우리의 24 비트 킬러 코딩을 이용한 뇌 회백질의 신경 다발의 방향성 가시화에 성공하였다. 우리가 구현한 방법에 의해 뇌의 주요 신경 다발들이 잘 가시화됨을 확인하였다.

• 대한정형도수물리치료학회지
• /
• 제26권1호
• /
• pp.47-54
• /
• 2020

Background: This study examined the changes in the leg muscle activities of 30 healthy subjects according to different pressures caused by isometric hip adduction during squat exercise. Methods: With their knee flexed at 60°, the subjects performed isometric hip adduction with a pressure of 0, 20, 40, and 60mmHg. Surface electromyography was used to measure their muscle activities. Changes in the activities of the muscles, such as the gluteus maximus (GMAX), Gluteus medius (GM), vastus medialis oblique (VMO), vastus lateralis oblique (VLO), biceps femoris (BF), and Tensor fasciae latae (TFL) muscles, were analyzed. Results: The activities of the VMO, VLO, GM, BF, and GMAX muscles were significantly different after exercise compared to that before exercise (p<.05). The activities of all the leg muscles measured were highest at a pressure of 60mmHg during isometric hip joint adduction. Conclusion: The pressure produced by isometric hip joint adduction during squat exercise increased the leg muscle activities of the subjects. These results will provide basic data on effective squat exercise to alleviate knee joint diseases.

• 대한수학회논문집
• /
• 제32권3호
• /
• pp.661-676
• /
• 2017

Let T be a bounded linear operator acting on a complex Hilbert space ${\mathfrak{H}}$. In this paper we introduce the class, denoted ${\mathcal{Q}}(A(k),m)$, of operators satisfying $T^{m{\ast}}(T^{\ast}{\mid}T{\mid}^{2k}T)^{1/(k+1)}T^m{\geq}T^{{\ast}m}{\mid}T{\mid}^2T^m$, where m is a positive integer and k is a positive real number and we prove basic structural properties of these operators. Using these results, we prove that if P is the Riesz idempotent for isolated point ${\lambda}$ of the spectrum of $T{\in}{\mathcal{Q}}(A(k),m)$, then P is self-adjoint, and we give a necessary and sufficient condition for $T{\otimes}S$ to be in ${\mathcal{Q}}(A(k),m)$ when T and S are both non-zero operators. Moreover, we characterize the quasinilpotent part $H_0(T-{\lambda})$ of class A(k) operator.

• 제어로봇시스템학회:학술대회논문집
• /
• 제어로봇시스템학회 1994년도 Proceedings of the Korea Automatic Control Conference, 9th (KACC) ; Taejeon, Korea; 17-20 Oct. 1994
• /
• pp.628-633
• /
• 1994

A new Riemannian geometric model for the controlled plant is proposed by imbedding the control vector space in the state space, so as to reduce the dimension of the model. This geometric model is derived by replacing the orthogonal straight coordinate axes on the state space of a linear system with the curvilinear coordinate axes. Therefore the integral manifold of the geometric model becomes homeomorphic to that of fictitious linear system. For the lower dimensional Riemannian geometric model, a nonlinear optimal regulator with a quadratic form performance index which contains the Riemannian metric tensor is designed. Since the integral manifold of the nonlinear regulator is determined to be homeomorphic to that of the linear regulator, it is expected that the basic properties of the linear regulator such as feedback structure, stability and robustness are to be reflected in those of the nonlinear regulator. To apply the above regulator theory to a real nonlinear plant, it is discussed how to distort the curvilinear coordinate axes on which a nonlinear plant behaves as a linear system. Consequently, a partial differential equation with respect to the homeomorphism is derived. Finally, the computational algorithm for the nonlinear optimal regulator is discussed and a numerical example is shown.

• International Journal of Aeronautical and Space Sciences
• /
• 제2권2호
• /
• pp.65-72
• /
• 2001

This paper demonstrates an explicit-implicit, finite element analysis for linear as well as nonlinear hygrothermal stress problems. Additional features, such as moisture diffusion equation, crack element and virtual crack extension(VCE) method for evaluating J-integral are implemented in this program. The Linear Elastic Fracture Mechanics(LEFM) Theory is employed to estimate the crack driving force under the transient condition for an existing crack. Pores in materials are assumed to be saturated with moisture in the liquid form at the room temperature, which may vaporize as the temperature increases. The vaporization effects on the crack driving force are also studied. The ideal gas equation is employed to estimate the thermodynamic pressure due to vaporization at each time step after solving basic nodal values. A set of field equations governing the time dependent response of porous media are derived from balance laws based on the mixture theory. Darcy's law is assumed for the fluid flow through the porous media. Perzyna's viscoplastic model incorporating the Von-Mises yield criterion are implemented. The Green-Naghdi stress rate is used for the invariant of stress tensor under superposed rigid body motion. Isotropic elements are used for the spatial discretization and an iterative scheme based on the full Newton-Raphson method is used for solving the nonlinear governing equations.

• Structural Engineering and Mechanics
• /
• 제13권2호
• /
• pp.135-154
• /
• 2002

This paper presents the convected material frame approach to study the nonlinear behavior of inelastic frame structures. The convected material frame approach is a modification of the co-rotational approximation by incorporating an adaptive convected material frame in the basic definition of the displacement vector and strain tensor. In the formulation, each discrete element is associated with a local coordinate system that rotates and translates with the element. For each load increment, the corresponding strain-displacement and nodal force-stress relationships are defined in the updated local coordinates, and based on the updated element geometry. The rigid body motion and deformation displacements are decoupled for each increment. This modified approach incorporates the geometrical nonlinearities through the continuous updating of the material frame geometry. A generalized nonlinear function is used to derive the inelastic constitutive relation and the kinematic hardening is considered. The equation of motion is integrated by an explicit procedure and it involves only vector assemblage and vector storage in the analysis by assuming a lumped mass matrix of diagonal form. Several numerical examples are demonstrated in close agreement with the solutions obtained by the ANSYS code. Numerical studies show that the proposed approach is capable of investigating large deflection of inelastic planar structures and providing an excellent numerical performance.

• Investigative Magnetic Resonance Imaging
• /
• 제16권2호
• /
• pp.115-123
• /
• 2012

목적: Tract-based spatial statistics와 화소기반 형태분석 방법을 이용하여 경도의 알츠하이머병 환자와 경도인 지장애 환자에서 확산텐서영상을 이용하여 뇌 백질의 이상을 평가하고자 하였다. 대상과 방법: 21명의 경도의 알츠하이머병 환자와 13명의 경도인지장애 환자, 그리고 16명의 건강한 노인군을 대상으로 확산텐서영상을 시행하였다. 각 참가자마다 분할 비등방도를 구하여 Tract-based spatial statistics를 이용하여 세 그룹간의 비교를 하였다. Tract-based spatial statistics 방법과의 비교를 위하여, 화소기반 형태분석 방법을 이용한 분석도 함께 시행하였다. 결과: Tract-based spatial statistics 분석결과, 경도의 알츠하이머 환자에서 건강한 노인군보다 방사관의 양측 전각과 우측 후각, 후시상방사, 우측 상세로다발, 뇌량체부, 우측 쐐기전소엽이랑에서 분할 비등방도가 유의하게 감소하였다. 화소기반 형태분석에서는 양측 갈고리다발, 좌측 부해마회의 백질, 우측 대상다발에서 추가적으로 분할 비등 방도가 감소되어 있었다. 경도의 알츠하이머 환자군과 경도인지장애 환자군의 비교, 경도인지장애 환자군과 건강한 노인군의 비교연구에서는 분할 비등방도의 유의한 차이는 없었다. 결론: Tract-based spatial statistics 분석결과, 경도의 알츠하이머 환자군에서 건강한 노인군에 비해 뇌 백질의 분할 비등방도가 여러 곳에서 감소되어 있었다. 화소기반 형태분석방법은 tract-based spatial statistics보다 더 많은 곳에서 뇌 백질의 이상을 보였다. 그러나, 인공물 또한 더 많이 관찰되었다.