• MALSORI
• /
• no.43
• /
• pp.151-161
• /
• 2002

This study presents performance analysis results of speaker adaptation for keyword spotting system. In this paper, we implemented MLLR (Maximum Likelihood Linear Regression) method on our middle size vocabulary keyword spotting system. This system was developed for directory services of universities and colleges. The experimental results show that speaker adaptation reduces the false alarm rate to 1/3 with the preservation of the mis-detection ratio. This improvement is achieved when speaker adaptation is applied to not only keyword models but also non-keyword models.

• Proceedings of the Korean Information Science Society Conference
• /
• 2012.06a
• /
• pp.7-9
• /
• 2012

CPU는 싱글 코어 구조에서 클록 속도를 높여 성능을 향상 시키려는 노력을 해왔으나 한계에 도달하자 하나의 칩에 코어를 여러 개 둔 멀티코어 형태로 발전하였다. CPU의 성능 향상을 위해 이제는 3D그래픽을 연산처리하기 위해 만들어진 GPU와 결합하기에 이르렀다. CPU와 GPU의 결합은 CPU간의 결합보다 훨씬 더 좋은 성능을 보였고 전력의 사용량도 더 적었으며 비용면에서도 경제적이라는 장점을 가지고 있다. 본 논문에서는 CPU와 GPU의 Heterogeneous multicore상에서 성능을 최적화하기 위해 기존의 병렬화 모델을 조합하고 최적화를 시도하였다. CPU상에서는 성능 향상을 위해 기존의 병렬 프로그램 모델인 SIMD와 공유메모리 병렬 프로그래밍 모델 그리고 메시지 패싱 병렬 프로그래밍 모델을 조합하는 실험을 했다. GPU에서는 CUDA를 최적화 하였다. 이렇게 CPU와 GPU를 최적화하고 조합하여 고성능 연산을 요구하는 어플리케이션을 위한 Heterogeneous multicore 성능 최적화 방법을 제안한다.

• Journal of Korea Society of Industrial Information Systems
• /
• v.4 no.4
• /
• pp.88-93
• /
• 1999

Several Iterative methods are considered, Gram-Schmidt algerian for thin orthogonalization and Lanczos methodfor a few extreme eigenvalues. For these methods, a variants of method is derived for which only one synchronization point per on iteration is required; that is one global communication in a message passing distributed-memory machine per one iteration is required The variant is called restructured method, and restructured method has better parallel properties to the conventional method.

• Proceedings of the Korean Society of Precision Engineering Conference
• /
• 2002.05a
• /
• pp.470-473
• /
• 2002

Most networks for automation are tuned to an expected traffic at their design stage. During their actual operations, however, the networks experience considerable changes in traffic from time to time. These traffic changes caused by common events like machine failure and production schedule change may adversely affect the network performance and, in turn, the performance of the connected devices. This paper presents communication delay model for Profibus token passing protocol, and introduces TTR selection methods to maintain a uniform level of network performance at all stations under changing network traffic.

• Proceedings of the KIEE Conference
• /
• 2006.07d
• /
• pp.1987-1988
• /
• 2006

본 논문에서는 필드버스를 이용한 지능형 전송기의 FOUNDATION 필드버스 응용계층에 관해 설계하였다. 필드버스 전송기에는 물리계층 데이터링크 계층 응용계층이 있는데 여기서는 응용 계층 중에서 FAS(Fieldbus Access Sublayer)에 대해서만 설명하고자 한다. FAS는 DLL(Data Link Layer) 계층과 응용 계층인 FMS(Fieldbus Message Specification)간에 인터페이스 기능을 수행한다. FMS에서 생성된 메시지가 DLL에서 제공하는 토큰 패싱 또는 스케줄 방식을 통해서 전송되기 위해서는 전송단과 수신단간에 통신관계가 설정되어야 하며, 이러한 기능이 FAS 계층에서 수행된다. FAS에서 통신관계를 설정하기 위한 서비스 종류는 VCR(Virtual Communication Relationship)에 의하여 결정된다. 하드웨어 설계 부분에 대해서는 본 논문에서 설명하지 않았다.

• Annual Conference of KIPS
• /
• 2002.11c
• /
• pp.1973-1976
• /
• 2002

본 논문에서는 유 무선 네트워크 환경에서의 자료 일치성 유지를 위한 동기화 프로토콜(SyncML1.1)을 검증한다. 프로토콜 검증에 관한 기존 연구에서는 약한 일관성을 갖는 동기화 프로토콜은 다루어지지 않았으며 정리 증명 혹은 모델 채킹에 대한 선택적 접근으로인해 증명 영역이 제한적이었다. 본 연구에서는 프로토콜의 의미를 정의하고 행위 및 시스템 구조, 메시지 패싱에 대한 개별적 모델을 단계적으로 설계하는 한편, 동기화 프로토콜에서의 안전성 및 상호 배제, 궁극성 속성은 시퀀트 계산법에서의 역방향 전략으로 증명되며, 모델 채킹 기술을 하나의 증명 규칙으로 사용했다. 본 연구는 두가지 부분에서 특징과 의미을 갖는다. 첫째, 정리증명과 모델 채킹이 혼합된 검증 방법을 사용했다. 둘째, 객체 제약 언어 및 정형 명세 언어의 사용을 통해 모델의 생성을 보다 단계적이고 정형적으로 유도했으며 이를 통해 검증의 신뢰성 및 품질을 높였다.

• 제어로봇시스템학회:학술대회논문집
• /
• 1989.10a
• /
• pp.524-529
• /
• 1989

In this paper, an approximate sojourn time distribution is obtained for cyclic service systems. We consider symmetric and limited service systems in which each queue has an infinite capacity. The combined service time is defined which consists of the frame service time and server waiting time that is approximated by two cases of the uniform and exponential distributions. The approximate sojourn time distribution is obtained from the Pollaczek-Khinchine formula where the combined service time is used for the service time in the M/G/I model. And some numerical examples are given to validate the suggested approximate analysis.

• Proceedings of the Korean Society of Precision Engineering Conference
• /
• 2003.06a
• /
• pp.181-184
• /
• 2003

Recently, the fieldbus becomes an indispensable component for many automated systems. In the fieldbus system, realtime data containing sensor values and control commands has a tendency to rapidly lose its value as time elapses after its creation. In order to deliver these data in time, the fieldbus network should be designed to have short delay compared to the maximum allowable delay. Because the communication delay is affected by performance parameters such as target rotation timer of token passing protocol, it is necessary to select proper parameter settings to satisfy the real-time requirement for communication delay. This paper presents the timer selection method for Profibus token passing networks using genetic algorithm (GA) to meet the delay requirements.

• Journal of Institute of Control, Robotics and Systems
• /
• v.9 no.12
• /
• pp.1055-1064
• /
• 2003

In many automated systems, such as manufacturing systems and process plants, a fieldbus is a very important component for the exchange of various and sometimes crucial information. Some of the information has a tendency to rapidly lose its value as time elapses after its creation. Such information or data is called real-time data that includes sensor values and control commands. In order to deliver these data in time, the fieldbus network should be tailored to have short delay with respect to the individual time limit of various data. Fine-tuning the network for a given traffic requires the knowledge on the relationship between the protocol parameters such as timer values and the performance measure such as network delay. This paper presents a mathematical performance model to calculate communication delays of the Profibus-FMS network when the timer value and the traffic characteristics are given.

• Proceedings of the KSME Conference
• /
• 2003.11a
• /
• pp.928-933
• /
• 2003

For simulation of three-dimensional turbulent flow with LES and DNS takes much time and expense with current available computing resources. It is nearly impossible to simulate turbulent flow with high Reynolds number. So, the emerging alternative is the Grid computing for needed computation power and working environment. In this study, the CFD code was parallelized to adapt it for the parallel computing under the Grid environment. In the first place, the Grid environment was built to connect the PC-Cluster facilities belong to the different institutions using communication network system. And CFD applications were calculated to check the performance of the parallel code developed for the Grid environment. Although it is a fundamental study, it brings about a important meaning as first step in research of the Grid.