• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.12 no.9
• /
• pp.4412-4428
• /
• 2018

The purpose of this work is to solve the problem of representing an entire video using Convolutional Neural Network (CNN) features for human action recognition. Recently, due to insufficient GPU memory, it has been difficult to take the whole video as the input of the CNN for end-to-end learning. A typical method is to use sampled video frames as inputs and corresponding labels as supervision. One major issue of this popular approach is that the local samples may not contain the information indicated by the global labels and sufficient motion information. To address this issue, we propose a binary hashing method to enhance the local feature extractors. First, we extract the local features and aggregate them into global features using maximum/minimum pooling. Second, we use the binary hashing method to capture the motion features. Finally, we concatenate the hashing features with global features using different normalization methods to train the classifier. Experimental results on the JHMDB and MPII-Cooking datasets show that, for these new local features, binary hashing mapping on the sparsely sampled features led to significant performance improvements.

• Spatial Information Research
• /
• v.12 no.4 s.31
• /
• pp.371-382
• /
• 2004

Continuously Operating Reference Stations (CORS) provides GPS measurements to support high accuracy CPS positioning. The CORS improves GPS positioning productivity by eliminating the requirement for GPS users to operate two receivers. Previously, this is achieved by providing data from CORS to users in post-mission mode. However, the efficiency of the CORS will be maximized by upgrading it in real-time operation using RTK-GPS surveying because users can obtain centimeter-level accuracy in real-time without operating their own reference stations. In this research, authors extracted the arbitrary point's coordinate which is using GPS CORS data, now served in RINEX FORMAT via Internet, with observation network of the existing triangulation and GPS CORS data. And then, RTK GPS was performed with this arbitrary point as reference station.

• Proceedings of the Korean Operations and Management Science Society Conference
• /
• 2002.05a
• /
• pp.285-292
• /
• 2002

A cluster tool for semiconductor manufacturing is an integrated device that consists of several single wafer processing modules and a wafer transport module based on a robot. The distributed module controllers are integrated by an inter-module communication network and coordinated by a centralized controller, called a cluster tool controller (CTC). Since the CTC monitors and coordinates the distributed complex module controllers for advanced process control, complex commuication messaging and services between the CTC and the module controllers are required. A SEMI standard, CTMC(Cluster Tool Module Communication), specifies application-level communication service requirements for inter-module communication. We propose the use of high-level fieldbuses, for instance. PROFIBUS-FMS, for implementing CTMC since the high-level fieldbuses are well suited for complex real-time distributed manufacturing control applications. We present a way of implementing CTMC using PROFIBUS-FMS as the communication enabler. We first propose improvements of a key object of CTMC for material transfer and the part transfer protocol to meet the functional requirements of modem advanced cluster tools. We also discuss mapping objects and services of CTMC to PROFIBUS-FMS communication objects and services. Finally, we explain how to implement the mappings.

• Journal of Institute of Control, Robotics and Systems
• /
• v.14 no.2
• /
• pp.168-177
• /
• 2008

Path finding is a key element in the navigation of a mobile robot. To find a path, robot should know their position exactly, since the position error exposes a robot to many dangerous conditions. It could make a robot move to a wrong direction so that it may have damage by collision by the surrounding obstacles. We propose a method obtaining an accurate robot position. The localization of a mobile robot in its working environment performs by using a vision system and Virtual Reality Modeling Language(VRML). The robot identifies landmarks located in the environment. An image processing and neural network pattern matching techniques have been applied to find location of the robot. After the self-positioning procedure, the 2-D scene of the vision is overlaid onto a VRML scene. This paper describes how to realize the self-positioning, and shows the overlay between the 2-D and VRML scenes. The suggested method defines a robot's path successfully. An experiment using the suggested algorithm apply to a mobile robot has been performed and the result shows a good path tracking.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.21 no.9
• /
• pp.2481-2493
• /
• 1996

A simple iterative design method is presented for the miniaturized dielectric filter with finite transimmission zeros. Circuit-network transformations are used to place transission zeros in the stop band. This method is simple and quick compared with conventional pole-synthesis methods. the design of the coupling EM patternof the filter needs the much computation time because its design is carried out through the field simulatior. In this paper, an efficient implementation technique of coupling circuit parameters is presented. This technique uses a linear mapping function between the circuit parameter domain and EM parameter domain to save the high computational time of EM simulator. A narrow band asymmetric filter with a transmission zero is designed and fiblicated through these technique in 1900 MHz band.

• ETRI Journal
• /
• v.27 no.4
• /
• pp.341-354
• /
• 2005

The adaptation of video according to the heterogeneous and dynamic resource constraints on networks and devices, as well as on user preferences, is a promising approach for universal access and consumption of video content. For optimal adaptation that satisfies the constraints while maximizing the utility that results from the adapted video, it is necessary to devise a systematic way of selecting an appropriate adaptation operation among multiple feasible choices. This paper presents a general conceptual framework that allows the formulation of various adaptations as constrained optimization problems by modeling the relations among feasible adaptation operations, constraints, and utilities. In particular, we present the feasibility of the framework by applying it to a use case of rate adaptation of MPEG-4 video with an explicit modeling of adaptation employing a combination of frame dropping and discrete cosine transform coefficient dropping, constraint, utility, and their mapping relations. Furthermore, we provide a description tool that describes the adaptation-constraint-utility relations as a functional form referred to as a utility function, which has been accepted as a part of the terminal and network quality of service tool in MPEG-21 Digital Item Adaptation (DIA).

• Publications of The Korean Astronomical Society
• /
• v.30 no.2
• /
• pp.633-636
• /
• 2015

KaVA (KVN and VERA Array) is a new combined VLBI array composed of KVN (Korean VLBI Network) and VERA (VLBI Exploration of Radio Astrometry). Here, we report the following two issues. (1) We review the initial results of imaging observations of M87 at 23 GHz following Niinuma et al. (2014). The KaVA images reveal extended outflows including complex substructures such as knots and limb-brightening, in agreement with previous VLBI observations. KaVA achieves a high dynamic range of ~1000, more than three times better than that achieved by VERA alone. (2) Based on subsequent observations and discussions led by the KaVA AGN SubWorking Group, we set monitoring observations of Sgr $A^{\ast}$ and M87 as our Key Science Project (hereafter KSP) because of the closeness and largeness of their central super-massive black holes. The main science goals of the KSP are (i) testing the magnetically-driven-jet paradigm by mapping velocity fields of the M87 jet, and (ii) obtaining tight constraints on physical properties of the radio emitting region in Sgr $A^{\ast}$. Towards KSP, we show the first preliminary images of M87 at 23 GHz and Sgr $A^{\ast}$ at 43 GHz with the bandwidth of 256 MHz.

• Journal of KIISE:Software and Applications
• /
• v.31 no.5
• /
• pp.537-542
• /
• 2004

Multi-Layer Perceptron network has been mainly applied to many practical problems because of its nonlinear mapping ability. However the generalization ability of MLP networks may be affected by the number of hidden nodes, the initial values of weights and the training errors. These factors, if improperly chosen, may result in poor generalization ability of MLP networks. It is important to identify these factors and their interaction in order to control effectively the generalization ability of MLP networks. In this paper, we have empirically identified the factors that affect the generalization ability of MLP networks, and compared their relative effects on the generalization performance for the conventional and visualized weight selecting methods using the controller box.

• Proceedings of the KIEE Conference
• /
• 2001.07d
• /
• pp.2762-2764
• /
• 2001

In this study, we introduce and investigate a class of dynamic perceptron architectures, discuss a comprehensive design methodology and carry out a series of numeric experiments. The proposed dynamic perceptron architectures are called as Polynomial Neural Networks(PNN). PNN is a flexible neural architecture whose topology is developed through learning. In particular, the number of layers of the PNN is not fixed in advance but is generated on the fly. In this sense, PNN is a self-organizing network. PNN has two kinds of networks, Polynomial Neuron(FPN)-based and Fuzzy Polynomial Neuron(FPN)-based networks, according to a polynomial structure. The essence of the design procedure of PN-based Self-organizing Polynomial Neural Networks(SOPNN) dwells on the Group Method of Data Handling (GMDH) [1]. Each node of the SOPNN exhibits a high level of flexibility and realizes a polynomial type of mapping (linear, quadratic, and cubic) between input and output variables. FPN-based SOPNN dwells on the ideas of fuzzy rule-based computing and neural networks. Simulations involve a series of synthetic as well as experimental data used across various neurofuzzy systems. A detailed comparative analysis is included as well.

• Proceedings of the KIEE Conference
• /
• 2001.07d
• /
• pp.1964-1966
• /
• 2001

Active Noise control(ANC) is rapidly becoming the most effective way to reduce noises that can otherwise be very difficult and expensive to control. This research presents ANFIS (Adaptive Network Fuzzy Inference System) controller for adaptively noise cancelling in a duct. ANC system generates secondary control sound pressure with same amplitude and with opposite phase as noise to be eliminated. ANFIS controller is trained to optimize its parameters for adaptively cancelling noise. That is ANFIS train its parameters by gradient descent and LSE method so called hybrid method. This paper present ANFIS in active noise control which provides an improvement convergence speed and limitation of linearity condition. It can model nonlinear functions of arbitrary complexity and ANFIS can construct an input-ouput mapping based on both human knowledge in the form of Takagi and Sugeno's fuzzy if-then rules and stipulated input-output data pairs. This paper also shows that the proposed ANFIS active noise control system successfully cancelled noise.