• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2019.11a
• /
• pp.137-139
• /
• 2019

본 논문은 다양한 음성 특징과 텍스트를 이용한 멀티 모드 순환신경망 네트워크를 사용하여 음성을 통한 범주형(categorical) 분류 방법과 Arousal-Valence(AV) 도메인에서의 분류방법을 통해 감정인식 결과를 제시한다. 본 연구에서는 음성 특징으로는 MFCC, Energy, Velocity, Acceleration, Prosody 및 Mel Spectrogram 등의 다양한 특징들의 조합을 이용하였고 이에 해당하는 텍스트 정보를 순환신경망 기반 네트워크를 통해 융합하여 범주형 분류 방법과 과 AV 도메인에서의 분류 방법을 이용해 감정을 이산적으로 분류하였다. 실험 결과, 음성 특징의 조합으로 MFCC Energy, Velocity, Acceleration 각 13 차원과 35 차원의 Prosody 의 조합을 사용하였을 때 범주형 분류 방법에서는 75%로 다른 특징 조합들 보다 높은 결과를 보였고 AV 도메인 에서도 같은 음성 특징의 조합이 Arousal 55.3%, Valence 53.1%로 각각 가장 높은 결과를 보였다.

• BMB Reports
• /
• v.42 no.12
• /
• pp.823-828
• /
• 2009

Techniques for analyzing protein-DNA interactions on a genome-wide scale have recently established regulatory roles for distal enhancers. However, the large sizes of higher eukaryotic genomes have made identification of these elements difficult. Information regarding sequence conservation, exon annotation and repetitive regions can be used to reduce the size of the search region. However, previously developed resources are inadequate for consolidating such information. CONVIRT is a web resource for the identification of transcription factor binding sites and also features comparative genomics. Genomic information on ortholog-independent conserved regions, exons, repeats and sequences is integrated into the virtual chromosome, and statistically over-represented single or combinations of transcription factor binding sites are sought. CONVIRT provides regulatory network analysis for several organisms with long promoter regions and permits inter-species genome alignments. CONVIRT is freely available at http://biosoft.kaist.ac.kr/convirt.

• Journal of Korean Library and Information Science Society
• /
• v.24
• /
• pp.449-484
• /
• 1996

The purpose of this study is to analyze the structural model of classification, to compare enumerative schemes with faceted schemes and finding the facted schems used in the Dewey Decimal Classification. The Structural model of classifications can be divided to enumerrative schemes and facted schemes(analytico-synthetic schemes). Enumerative schemes represent the various kind of subject in different ways and have to not only list all existing combinations of subjects but also any of potential subjects. Faceted schemes is the decomposition of concepts into all possible characteristics which are known as facets. DDC was originally created as enumerative, but over 20 editions has moved increasing towards synthetic features such as the add instructions, and the seven auxiliary tables. Many DDC class numbers found in the Schedules may be subdivied by another number(or part of it) that has been drawn from the Schedules or auxiliary tables. This process of subdividing class numbers is called number building and it can be done only when there are number building instruction in the Schedules or Tables.

• Proceedings of the KIEE Conference
• /
• 1987.07b
• /
• pp.1598-1601
• /
• 1987

This paper proposes a new set of topological features (primitives) for use with a syntactic recognizer for high-accuracy recognition of printed alphanumeric characters. The recognition is accomplished on nine character groups, where each group has different combinations of four feature points. A skeleton enhancement eliminating isolated points and smoothing irregular points is developed. The tree automata processed in parallel enables the realization of high-recognition speeds and font-type independent recognition. The proposed character recognition system is tested for alphanumeric character fonts of dot matrix printer and plotter using IBM-PC/XT.

• Speech Sciences
• /
• v.6
• /
• pp.119-143
• /
• 1999

In this paper, the development of English Intonational Phonology is introduced. The existing representation systems of intonation are largely divided into the American structuralist school and the British school, which describe intonation by means of 'levels' and 'configurations' respectively. Both representation systems have some theory-internal problems, however. As for the American school, there is no way to represent pitches much lower than the reference line, while the system of intonation in the British school is limited in that intonation is described in a phonetic impressionistic way rather than from a phonological perspective. Intonational Phonology, a real phonological approach, which has grown out of the basic assumptions of autosegmental-metrical(AM) theory has been suggested by Pierrehumbert(1980). In her approach, an intonational tune is made up of one or more pitch accents, followed by an obligatory phrase accent and an obligatory boundary tone, and interestingly 22 combinations are possible. Intonational Phonology has been revised from Beckman & Pierrehumbert(1986) in developing ToBI(Tones & Break Indices), a proposed standard for labelling prosodic features of digital speech databases in English.

• The Transactions of the Korean Institute of Electrical Engineers D
• /
• v.51 no.10
• /
• pp.446-452
• /
• 2002

For most of pattern recognition applications, it is required to correctly recognize patterns even if they have translation variations. In this paper, to achieve the goal of translation invariant pattern recognition, we propose a new generalized translation invariant second-order neural network using a constraint on the weights. The weight constraint is implemented using generalized translation invariant features which are accumulated sums of pixel combinations. Simulation results will be given to demonstrate that the proposed second-order neural network has the generalized translation invariant property.

• Korean Management Science Review
• /
• v.24 no.2
• /
• pp.57-72
• /
• 2007

Many Applications of scientific research have coupled with functional data signal clustering techniques to discover novel characteristics that can be used for the diagnoses of several issues. In this article we present an interpretable multi-scale cluster hierarchy framework for clustering functional data using its multi-aspect frequency information. The suggested method focuses on how to effectively select transformed features/variables in unsupervised manner so that finally reduce the data dimension and achieve the multi-purposed clustering. Specially, we apply our suggested method to mutual fund returns and make superior-performing funds group based on different aspects such as global patterns, seasonal variations, levels of noise, and their combinations. To promise our method producing a quality cluster hierarchy, we give some empirical results under the simulation study and a set of real life data. This research will contribute to financial market analysis and flexibly fit to other research fields with clustering purposes.

• Journal of the Korea Society of Computer and Information
• /
• v.24 no.9
• /
• pp.69-75
• /
• 2019

We propose a method to analyze user requirements to design a virtual reality-based risk management system. This paper presents surveys, interviews, prototype evaluation methods, and implementation process. Architectural heritage is easily exposed to natural and artificial dangers caused by various material combinations and structural features. So, risk management of cultural heritage plays a key role in preserving and managing cultural heritage. However, risk management has been carried out through empirical methods using distributed data. This study analyzes user requirements for designing functions and interfaces of VR-based risk management system and evaluates prototypes to overcome the above problems. As a result, most heritage managers wanted a system function to support risk analysis and response. They also found that they prefer 2D information such as existing drawings and photos rather than 3D information. The results of the user requirements analysis derived from this study will be used to create risk management applications.

• Journal of Power Electronics
• /
• v.19 no.4
• /
• pp.944-955
• /
• 2019

In this paper, the three-level T-type neutral-point-clamped indirect matrix converter topology and the relative space vector modulation methods are introduced to improve the voltage transfer ratio and output voltage performance. The presented converter topology is based on combinations of cascaded-rectifier and three-level T-type neutral-point-clamp inverter. It can overcome the limitation of voltage transfer ratio of the conventional matrix converter and the high voltage rating of power switches of conventional matrix converter. Two SVPWM strategies for proposed converter are described in this paper to achieve the advantages features such as: sinusoidal input/output currents and three-level output voltage waveforms. Results from Psim 9.0 software simulation are provided to confirm the theoretical analysis. Hence, a laboratory prototype was implemented, and the experimental results are shown to validate the simulation results and to verify the effectiveness of the proposed topology and modulation strategies.

• Proceedings of the PSK Conference
• /
• 2002.10a
• /
• pp.367.1-367.1
• /
• 2002

Flavo.no.ids are a diverse group of phytqchemicals that are produced by various plants in high quantities. Dietary flavonoids in edible plants can be further subdivided into. several structural groups. The large number of compounds arises from various combinations of multiple hydroxyl and methoxyl groups substituting the basic flavonoid skeleton. The chemopreentive activity of flavonoids is dependent on their structural features. (omitted)