• The Journal of the Acoustical Society of Korea
• /
• v.34 no.3
• /
• pp.234-239
• /
• 2015

Non-Negative Matrix Factorization (NMF) is a method for updating dictionary and gain in alternating manner. Due to ease of implementation and intuitive interpretation, NMF is widely used to detect and separate overlapping sound events. However, NMF that utilizes non-negativity constraints generates parts-based representation and this distinct property leads to a dictionary containing fragmented acoustic events. As a result, the presence of shared basis results in performance degradation in both separation and detection tasks of overlapping sound events. In this paper, we propose a new method that utilizes K-Singular Value Decomposition (K-SVD) based dictionary to address and mitigate the part-based representation issue during the dictionary learning step. Subsequently, we calculate the gain using NMF in sound event detection step. We evaluate and confirm that overlapping sound event detection performance of the proposed method is better than the conventional method that utilizes NMF based dictionary.

• IE interfaces
• /
• v.25 no.1
• /
• pp.40-51
• /
• 2012

Today, many manufacturing companies realize that collaboration is crucial for their survival. Especially, in the perspective of quality, the importance of collaboration is emphasized because economic loss increases exponentially while defective parts go through the process in supply chain. However, the manufacturing companies are facing two main difficulties in implementing collaborative relationships with their suppliers. First, it is difficult for the suppliers to produce reliable products due to their obsolete facilities. The problem gets worse for second- or third-tire vendors. Second, the companies experience the lack of universally understandable set of terminology and effective methodologies for knowledge representation. Ontology is one of the best approaches to expressing and processing a domain knowledge. In this paper, we propose the manufacturing ontology-based quality prediction framework to represent and share the knowledge of industrial environment and to predict product quality in manufacturing processes. In addition, we develop the ontology-based quality prediction system based on the proposed framework. We carried out a series of experiments for an injection molding process at an automotive part supplier. The experimental results demonstrated that the proposed framework and system can be successfully applicable in manufacturing industry.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.11 no.5
• /
• pp.2607-2627
• /
• 2017

Nonnegative matrix factorization (NMF) has received considerable attention due to its effectiveness of reducing high dimensional data and importance of producing a parts-based image representation. Most of existing NMF variants attempt to address the assertion that the observed data distribute on a nonlinear low-dimensional manifold. However, recent research results showed that not only the observed data but also the features lie on the low-dimensional manifolds. In addition, a few hard priori label information is available and thus helps to uncover the intrinsic geometrical and discriminative structures of the data space. Motivated by the two aspects above mentioned, we propose a novel algorithm to enhance the effectiveness of image representation, called Dual graph-regularized Constrained Nonnegative Matrix Factorization (DCNMF). The underlying philosophy of the proposed method is that it not only considers the geometric structures of the data manifold and the feature manifold simultaneously, but also mines valuable information from a few known labeled examples. These schemes will improve the performance of image representation and thus enhance the effectiveness of image classification. Extensive experiments on common benchmarks demonstrated that DCNMF has its superiority in image classification compared with state-of-the-art methods.

• Proceedings of the IEEK Conference
• /
• 2008.06a
• /
• pp.723-724
• /
• 2008

In this paper, we explore face recognition technology for embedded system. We develop an algorithm suitable for systems under ubiquitous environment. The basic requirements includes appropriate data format and ratio of feature data to achieve efficiency of algorithm. Our experiment presents a face recognition technique for handheld devices. The essential parts for face recognition based on embedded system includes; integer representation from floating point calculation and optimization for memory management.

• The Mathematical Education
• /
• v.57 no.1
• /
• pp.37-54
• /
• 2018

Despite the significance of fraction in elementary mathematics education, it is not easy to teach it meaningfully in connection with real life in Korea. This study aims to investigate and analyze 3rd grade students' understanding on unit fraction concepts and on comparison of unit fractions and to identify the parts which need to be supplemented in relation to unit fraction. For these purposes, I reviewed previous studies and extracted chapters which cover unit fractions in elementary mathematics textbooks based on 2009 revised curriculums and analyzed teaching contexts and visual representations of unit fractions. From this point of view, I constructed a test which consists of three problems based on Chval et al(2013) to investigate students' understanding on unit fraction. To apply this test, I selected forty-one 3rd grade students and examined that students' aspects of understanding on unit fraction. The results were analyzed both qualitatively and quantitatively. In this study, I present the analysis results and provide implications and some didactical suggestions for teaching contexts and visual representations of unit fraction based on the discussion.

• Proceedings of the Korean Society of Precision Engineering Conference
• /
• 1992.04a
• /
• pp.297-300
• /
• 1992

For the most part of machine design, assembly drawing isaffected byits inserted parts. When a part is inserted in assembly, alreadyd stablished parting line is modified for reasonable representation. for example, break or trim fuction between the lines gives more recognizable drqwing. Most of parts is composed of such that point, line, arc, circle and block. If the part is composed of block, it is very difficult to operate because its attribution is unknown before exploding, but, It is very easy to operate except block. In this paper, it is required that basic feature recognition rules are caused by knowledge-based. This system which shows a good modify tool using AutoCAD system and ADS (Autocad Development System).

• Journal of the Computational Structural Engineering Institute of Korea
• /
• v.25 no.4
• /
• pp.307-314
• /
• 2012

IFC-based representation method of part library for superstructure module of modular steel bridge is proposed. The library is capable of efficiently offering and exchanging part information in process of manufacture, assembly, design, and construction of modular steel bridge. Entities, representing physical part information in IFC model, are matched semantically with parts of the superstructure module for representation of part information with IFC model. Either types of matched entities are applied in order to verify the role of each part, or new types are defined as a user-defined types. In addition, assembly system has been classified and defined into 4 levels of LoD(Level of Detail) to provide appropriate part information efficiently from the part library in each step of the process. Then, new property is defined for representing the LoD information with IFC Model. Finally, IFC-based test library of modular steel bridge is generated by applying the matched entities and entity types to the actual the superstructure module of modular steel bridge.

• The Korean Journal of Archival Studies
• /
• no.59
• /
• pp.5-46
• /
• 2019

This study aims to explore the concepts and nature of records in digital environments based on analyzing recent theoretical studies of records. This is not intended to defining records for management in a specific environment. This study is divided into two parts. Firstly, it classifies the definitions of records in three perspectives; evidence, information, and representation of activities, and explores the concepts and theoretical backgrounds underlying each definition. Secondly, it articulates some implications for understanding the concept and nature of records in digital environments by four categories such as i) fixity, ii) materiality, iii) activity and context, and iv) trustworthiness, which were derived from the definition of "persistent representation of activities" by Geoffrey Yeo.

• Korean Journal of Computational Design and Engineering
• /
• v.1 no.2
• /
• pp.133-149
• /
• 1996

In order to reduce the trial-and-errors in design and production of injection molded plastic parts, there has been much research effort not only on CAE systems which simulate the injection molding process, but also on CAD systems which support initial design and re-design of plastic parts and their molds. The CAD systems and CAE systems have been developed independently with being built on different basis. That is, CAD systems manipulate the part shapes and the design features in a complete solid model, while CAE systems work on shell meshes generated on the abstract sheet model or medial surface of the part. Therefore, it is required to support the two types of geometric models and feature information in one environment to integrate CAD and CAE systems for accelerating the design speed. A feature-based non-manifold geometric modeling system has been developed to provide an integrated environment for design and analysis of injection molding products. In this system, the geometric models for CAD and CAE systems are represented by a non-manifold boundary representation and they are merged into a single geometric model. The suitable form of geometric model for any application can be extracted from this model. In addition, the feature deletion and interaction problem of the feature-based design system has been solved clearly by introducing the non-manifold Boolean operation based on 'merge and selection' algorithm. The sheet modeling capabilities were also developed for easy modeling of thin plastic parts.

• Proceedings of the Korean Society of Precision Engineering Conference
• /
• 2005.10a
• /
• pp.119-123
• /
• 2005

Mold design is a knowledge-intensive decision making process where product designer, injection molding engineer as well as mold designer affect each other. Representation and management of design knowledge is a prerequisite for an intelligent design system, which aims to guide and support designer to carry out design activity in more efficient way by avoiding or minimizing unnecessary trial and errors. This paper discusses the issues in knowledge-based mold design, and describes the structure of a knowledge-based mold design system fur parts with micro features under development.