• Journal of Korea Water Resources Association
• /
• v.46 no.6
• /
• pp.585-596
• /
• 2013

The objective of this study is to evaluate the hybrid downscaling method combined Step-Wise Scaling (SWS) method with Regional Climate Model (RCM) simulation data for climate change impact study on hydrology area. The SWS method is divided by 3 categories (extreme event, dry event and the others). The extreme events, wet-dry days and the others are corrected by using regression method, quantile mapping method, mean & variance scaling method. The application and evaluation of SWS method with 3 existing and popular statistical techniques (linear scaling method, quantile mapping method and weather generator method) were performed at the 61 weather stations. At the results, the accuracy of corrected simulation data by using SWS are higher than existing 3 statistical techniques. It is expected that the usability of SWS method will grow up on climate change study when the use of RCM simulation data are increasing.

• Proceedings of the Korean Society of Computer Information Conference
• /
• 2022.07a
• /
• pp.251-254
• /
• 2022

본 논문은 XGBoost 알고리즘 기반의 침입탐지모델의 성능을 향상하기 위한 스케일링(scaling) 및 특성선택(feature selection) 기법을 제안한다. 머신러닝 모델 개발 중 전처리 단계에서 스케일링 및 특성선택을 수행하면 데이터세트의 조건수가 감소하여 모델의 성능을 향상할 수 있다. 각 과정별로 다양한 기법이 있지만 기존의 연구에서는 이러한 기법들을 적용한 결과를 비교·분석하지 않고 특정 기법을 적용한 결과만을 나열하였고 스케일링 및 특성선택에 대해 최적의 조합은 제시하지 못하였다. 따라서 본 논문에서는 다양한 전처리 기법들의 적용결과를 비교하고 최적의 조합을 제안한다. 또한 기존의 연구들이 특정 데이터세트에만 적용 가능한 전처리 기법을 제안하는데 비해 본 논문은 다양한 데이터세트에 대해 공통적으로 적용 가능한 전처리 기법을 제안함으로써 제안 기법의 범용성과 실세계 적용 가능성을 증명한다.

• Proceedings of the Acoustical Society of Korea Conference
• /
• 1998.06e
• /
• pp.319.2-322
• /
• 1998

본 논문에서는 반음절 단위 HMM을 이용한 연속 숫자 음성인식 시스템의 2단계로 이루어지는 화자 적응 알고리즘을 수행하였다. 음성인식 시스템에서 사용되는 훈련데이터의 양이 많더라도 발성속도, 발성크기 등의 화자 발성 습관에 따라 화자독립 음성인식 시스템에서는 많은 문제점들이 발생하게 된다. 불특정 화자를 대상으로 한 음성 인식에 있어서 개인차에 의한 변동을 대처하는 방법으로 유효한 음향적 특성을 추출하기 위해 스펙트럼의 동적인(Dynamic) 특성을 주로 이용하고 있다. 따라서 본 논문에서는 화자 적을 기법의 하나인 frequency warped spectral matching 방법을 연속숫자 음성 인식시스템에 적용하였으며, 이때 인식에 의한 적절한 화자별 스케일링 계수 선정 방법을 수행하여 오인식률이 감소함을 확인하였다.

• Journal of the Korea Society of Computer and Information
• /
• v.18 no.7
• /
• pp.11-21
• /
• 2013

Recently operating a large scale computing resource like a data center becomes easier because of the virtualization technology that virtualize servers and enable flexible resource provision. The most of public cloud services provides automatic scaling in the form of scale-in or scale-out and these scaling approaches works well to satisfy the service level agreement (SLA) of users. However, a novel scaling approach is required to operate private clouds that has smaller amount of computing resources than vast resources of public clouds. In this paper, we propose a hybrid server scaling architecture and related algorithms using both scale-in and scale-out to achieve higher resource utilization rate for private clouds. We uses dynamic resource allocation and live migration to run our proposed algorithm. Our propose system aims to provide a fine-grain resource scaling by steps. Thus private cloud systems are able to keep stable service and to reduce server management cost by optimizing server utilization. The experiment results show that our proposed approach performs better in resource utilization than the scale-out approach based on the number of users.

• Journal of KIISE:Computer Systems and Theory
• /
• v.31 no.12
• /
• pp.726-734
• /
• 2004

This paper presents a facial animation method that enables the user to select a sequence of facial frames from the facial expression space, whose level of details the user can select hierarchically Our system creates the facial expression space from about 2400 captured facial frames. To represent the state of each expression, we use the distance matrix that represents the distance between pairs of feature points on the face. The shortest trajectories are found by dynamic programming. The space of facial expressions is multidimensional. To navigate this space, we visualize the space of expressions in 2D space by using the multidimensional scaling(MDS). But because there are too many facial expressions to select from, the user faces difficulty in navigating the space. So, we visualize the space hierarchically. To partition the space into a hierarchy of subspaces, we use fuzzy clustering. In the beginning, the system creates about 10 clusters from the space of 2400 facial expressions. Every tine the level increases, the system doubles the number of clusters. The cluster centers are displayed on 2D screen and are used as candidate key frames for key frame animation. The user selects new key frames along the navigation path of the previous level. At the maximum level, the user completes key frame specification. We let animators use the system to create example animations, and evaluate the system based on the results.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.35 no.2A
• /
• pp.209-215
• /
• 2010

This paper describes a design of variable-length FFT/IFFT processor (VL_FCore) for OFDM-based multi-standard communication systems. The VL_FCore adopts in-place single-memory architecture, and uses a hybrid structure of radix-4 and radix-2 DIF algorithms to accommodate various FFT lengths in the range of $N=64{\times}2^k\;(0{\leq}k{\leq}7)$. To achieve both memory size reduction and the improved SQNR, a two-step conditional scaling technique is devised, which conditionally scales the intermediate results of each computational stage. The performance analysis results show that the average SQNR's of 64~8,192-point FFT's are over 60-dB. The VL_FCore synthesized with a $0.35-{\mu}m$ CMOS cell library has 23,000 gates and 32 Kbytes memory, and it can operate with 75-MHz@3.3-V clock. The 64-point and 8,192-point FFT's can be computed in $2.25-{\mu}s$ and $762.7-{\mu}s$, respectively, thus it satisfies the specifications of various OFDM-based systems.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2009.10a
• /
• pp.393-396
• /
• 2009

This paper describes a design of variable-length FFT/IFFT processor for OFDM-based communication systems. The designed FFT/IFFT processor adopts the in-place single-memory architecture, and uses a hybrid structure of radix-4 and radix-2 DIF algorithms to accommodate FFT lengths of $N=64{\times}2^k$ ($0{\leq}k{\leq}7$). To achieve both memory size reduction and the improved SQNR, a two-step conditional scaling technique is devised, which conditionally scales the intermediate results of each computational stage. The performance analysis results show that the average SQNR's of 64~8,192-point FFT's are over 60-dB. The processor synthesized with a $0.35-{\mu}m$ CMOS cell library can operate with 75-MHz@3.3-V clock, and 64-point and 8,192-point FFT's can be computed in $2.55-{\mu}s$ and $762.7-{\mu}s$, respectively, thus it satisfies the specifications of wireless LAN, DMB, and DVB systems.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.8 no.2
• /
• pp.430-439
• /
• 2004

This paper describes a design of 8192/2048-point FFT/IFFT processor (CFFT8k2k), which performs multi-carrier modulation/demodulation in OFDM-based DVB-T receiver. Since a large size FFT requires a large buffer memory, two design techniques are considered to achieve memory-efficient implementation of 8192-point FFT/IFFT. A hybrid structure, which is composed of radix-4 single-path delay feedback (R4SDF) and radix-4 single-path delay commutator (R4SDC), reduces its memory by 20% compared to R4SDC structure. In addition, a memory reduction of about 24% is achieved by a novel two-step convergent block floating-point scaling. As a result, it requires only 57% of memory used in conventional design, reducing chip area and power consumption. The CFFT8k2k core is designed in Verilog-HDL, and has about 102,000 Bates, RAM of 292k bits, and ROM of 39k bits. Using gate-level netlist with SDF which is synthesized using a $0.25-{\um}m$ CMOS library, timing simulation show that it can safely operate with 50-MHz clock at 2.5-V supply, resulting that a 8192-point FFT/IFFT can be computed every 164-${\mu}\textrm{s}$. The functionality of the core is fully verified by FPGA implementation, and the average SQNR of 60-㏈ is achieved.

• Journal of the Institute of Electronics Engineers of Korea CI
• /
• v.45 no.1
• /
• pp.25-36
• /
• 2008

Eye localization means localization of the center of the pupils, and is necessary for face recognition and related applications. Most of eye localization methods reported so far still need to be improved about robustness as well as precision for successful applications. In this paper, we propose a robust eye localization method using multi-scale Gabor feature vectors without big computational burden. The eye localization method using Gabor feature vectors is already employed in fuck as EBGM, but the method employed in EBGM is known not to be robust with respect to initial values, illumination, and pose, and may need extensive search range for achieving the required performance, which may cause big computational burden. The proposed method utilizes multi-scale approach. The proposed method first tries to localize eyes in the lower resolution face image by utilizing Gabor Jet similarity between Gabor feature vector at an estimated initial eye coordinates and the Gabor feature vectors in the eye model of the corresponding scale. Then the method localizes eyes in the next scale resolution face image in the same way but with initial eye points estimated from the eye coordinates localized in the lower resolution images. After repeating this process in the same way recursively, the proposed method funally localizes eyes in the original resolution face image. Also, the proposed method provides an effective illumination normalization to make the proposed multi-scale approach more robust to illumination, and additionally applies the illumination normalization technique in the preprocessing stage of the multi-scale approach so that the proposed method enhances the eye detection success rate. Experiment results verify that the proposed eye localization method improves the precision rate without causing big computational overhead compared to other eye localization methods reported in the previous researches and is robust to the variation of post: and illumination.