• Annual Conference on Human and Language Technology
• /
• 2023.10a
• /
• pp.711-715
• /
• 2023

본 연구는 대부분의 언어 모델이 사용하고 있는 서브워드 토큰화 과정을 거치지 않고, 바이트 단위의 인코딩을 그대로 다룰 수 있는 토큰-프리 사전학습 언어모델에 대한 것이다. 토큰-프리 언어모델은 명시적인 미등록어 토큰이 존재하지 않고, 전 처리 과정이 단순하며 다양한 언어 및 표현 체계에 대응할 수 있는 장점이 있다. 하지만 관련 연구가 미흡, 서브워드 모델에 대비해 학습이 어렵고 낮은 성능이 보고되어 왔다. 본 연구에서는 한국어를 중심으로 토큰-프리 언어 이해-생성 모델을 사전 학습 후, 서브워드 기반 모델과 비교하여 가능성을 살펴본다. 또한, 토큰 프리 언어모델에서 지적되는 과도한 연산량을 감소시킬 수 있는 그래디언트 기반 서브워드 토크나이저를 적용, 처리 속도를 학습 2.7배, 추론 1.46배 개선하였다.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.35 no.7A
• /
• pp.632-638
• /
• 2010

In this paper, we propose a fixed-complexity sphere encoder (FSE) for multi-user MIMO (MU-MIMO) systems. The Proposed FSE accomplishes a scalable tradeoff between performance and complexity. Also, because it has a parallel tree-search structure, the proposed encoder can be easily pipelined, leading to a tremendous reduction in the precoding latency. The complexity of the proposed encoder is also analyzed, and we propose two techniques that reduce it. Simulation and analytical results demonstrate that in a $4\times4$ MU-MIMO system, the complexity of the proposed FSE is 16% that of the conventional QRD-M encoder (QRDM-E). Also, the encoding throughput of the proposed endoder is 7.5 times that of the QRDM-E with tolerable degradation in the BER performance, while achieving the optimum diversity order.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.17 no.6
• /
• pp.69-78
• /
• 2017

Shannon of the 5G smartphone and Fourier of the signal processing meet in the sampling theorem (2 times the highest frequency 1). In this paper, the initial Shannon Theorem finds the Shannon capacity at the point-to-point, but the 5G shows on the Relay channel that the technology has evolved into Multi Point MIMO. Fourier transforms are signal processing with fixed parameters. We analyzed the performance by proposing a 2N-1 multivariate Fourier-Jacket transform in the multimedia age. In this study, the authors tackle this signal processing complexity issue by proposing a Jacket-based fast method for reducing the precoding/decoding complexity in terms of time computation. Jacket transforms have shown to find applications in signal processing and coding theory. Jacket transforms are defined to be $n{\times}n$ matrices $A=(a_{jk})$ over a field F with the property $AA^{\dot{+}}=nl_n$, where $A^{\dot{+}}$ is the transpose matrix of the element-wise inverse of A, that is, $A^{\dot{+}}=(a^{-1}_{kj})$, which generalise Hadamard transforms and centre weighted Hadamard transforms. In particular, exploiting the Jacket transform properties, the authors propose a new eigenvalue decomposition (EVD) method with application in precoding and decoding of distributive multi-input multi-output channels in relay-based DF cooperative wireless networks in which the transmission is based on using single-symbol decodable space-time block codes. The authors show that the proposed Jacket-based method of EVD has significant reduction in its computational time as compared to the conventional-based EVD method. Performance in terms of computational time reduction is evaluated quantitatively through mathematical analysis and numerical results.

• The Transactions of the Korea Information Processing Society
• /
• v.4 no.11
• /
• pp.2671-2678
• /
• 1997

The MHEG-5 standard defines representation of multimedia and hypermedia information. The MHEG-5 can make it possible to interchange and present information between heterogeneous networked environment. Also, it can provide real-time user interaction. MHEG-5 object generating is the essential requirements for the MHEG-5 applications, In this paper, we designed encoding/decoding class library for the MHEG-5 objects. We can use it for the interchange on the network. The designed class library can provide base technology for the MHEG-5 based distributed multimedia and hypermedia services on the networked environment.

• Proceedings of the Korean Information Science Society Conference
• /
• 2001.10b
• /
• pp.211-213
• /
• 2001

본 논문에서는 서버와 클라이언트로 구성되는 분산환경에서 MPEG-4 데이터의 동기화를 제어하기 위한 방법과 그러한 방법으로 실제 구현이 된 MPEG-4 재생기를 소개한다. 데이터를 전송할 때 멀티 채널을 사용할 경우 얻을 수 있는 장점을 이용하기 위해서 서버측에서 MPEG-4 데이터를 디먹싱하여 비디오 부분과 오디오 부분을 분리해서 두 개의 채널로 클라이언트로 전송하며. 클라이언트는 이 데이터를 디코딩 타임 스템프(DTS)를 이용해서 화면상에 동기화를 구현하게 된다. MPEG 표준에서는 디코딩 타임 스템프(DTS)와 프리젠테이션 타임 스템프(PTS)를 정의하고 있는데, 본 논문에서는 DTS만으로 동기화를 구현을 한으로써 사용자 인터페이스 상의 재생 효율을 보장을 하면서 동시에 자원의 사용을 줄여 시스템 오버헤드를 줄일 수 있다. 이러한 방식은 무선환경에서의 멀티미디어 재생에 응용할 수 있을 것이다.

• Proceedings of the Korea Society for Simulation Conference
• /
• 2002.11a
• /
• pp.73-77
• /
• 2002

이 논문은 빠른 ASIP(application specific instruction processor) 시뮬레이션을 위한 재적응성을 가진 컴파일드 시뮬레이션 기법에 대해 이야기 한다. 다양한 응용분야에서의 설계 요구사항을 충족시키는 ASIP의 빠른 개발을 위해서, 건전한 설계 방법론 및 고성능의 시뮬레이터가 필요하다. 본 논문에서는 HiX$R^2$라는 ADL(architecture description language)을 이용하여 인스트럭션 수준에서 컴파일드 시뮬레이터를 자동 생성하였다. 컴파일드 시뮬레이션은 시뮬레이션 수행 시 반복되는 인스트럭션 페칭 및 디코딩 부분을 시뮬레이션 런-타임 이전에 미리 수행함으로서 일반적으로 사용되는 인터프리티브 시뮬레이션에 비하여 큰 성능향상을 얻을 수 있다. HiX$R^2$에 기반 한 컴파일드 시뮬레이션은 ARM9 프로세서와 CalmRISC32 프로세서 예제들로 수행하였고, 결과로서 인터프리티브 방식에 비해 150배 이상의 성능향상이 있었다.

• 한국정보통신설비학회:학술대회논문집
• /
• 2008.08a
• /
• pp.12-16
• /
• 2008

In this paper, we study precoding techniques for co-channel interference suppression in multiuser MIMO systems. DPC is optimal techniques to achieve the system capacity of multiuser MIMO downlink channels. DPC is not proper in practical wireless systems because complexity is very high. So block diagonal precoding for multiuser MIMO downlink channel is studied. The block diagonal precoding is used to suppress co-channel interference between multiuser. Block diagonal precoding method, whose complexity is reduced by modified null space operation, change multiuser MIMO channel to multiple single-user MIMO channel. We also use V-BLAST decoder in receiver. V-BLAST decoder can achieve increased system capacity in proportion to the number of users. We show improved system performance by using computer simulation.

• Journal of Broadcast Engineering
• /
• v.22 no.6
• /
• pp.755-764
• /
• 2017

In this paper, we design transmitter and receiver structures for a $2{\times}2$ multiple-input multiple-output (MIMO) in ATSC 3.0 systems and analyze the performance of the $2{\times}2$ MIMO system. In the ATSC 3.0 MIMO systems, spatial diversity and multiplexing gains can be obtained using the spatial demultiplexer and precoder. In this paper, we present the structures of the transmitter and receiver for ATSC 3.0 MIMO systems. Also, we present performance results of the $2{\times}2$ MIMO system through computer simulations.

• Journal of Korea Society of Digital Industry and Information Management
• /
• v.8 no.4
• /
• pp.115-120
• /
• 2012

In multi-user multi-input multi-output (MU-MIMO) system, zero forcing beamforming (ZFB) is regarded as a realistic solution for transmitting scheme due to its low complexity and simple structure. However, ZFB shows a significant performance degradation when channel matrix has large condition number. In this case, the largest singular value of the channel inversion matrix has a dominant effect on transmit power. In this paper, we propose a perturbation method for reducing an effect of the dominant singular value. In the proposed algorithm, channel inverse matrix is first decomposed by SVD for the transmit signal to be expressed as a combination of singular vectors. Then, the transmit signal is perturbed to reduce the coefficient of the singular vector corresponding to the largest singular value. When a number of transmit antennas is 4, the simulation results of this paper shows that the proposed method shows 8dB performance enhancement at 10-3 uncoded bit error rate (BER) compared with conventional ZFB. Also, the simulation results show that the proposed method provides a comparable performance to Tomlinson-Harashima Precoding (THP) with much lower complexity.

• The KIPS Transactions:PartD
• /
• v.11D no.6
• /
• pp.1239-1246
• /
• 2004

The integration and union of regional digital map and vehicle location data in e-Logistics which means virtual service architecture based on the Internet among the logistics companies is indispensable clause in order to satisfy the visibility characteristics. Most conventional vehicle monitoring and controlling systems were constructed with individual and mutual exclusive architecture so that there are required lots of cost for sharing and integrating the information among them. In this paper, we design and implement an component-based open system which represent moving object location information as well as digital map using the GML(Geography Markup Language - international standard encoding based on XML). We also test proposed system on the delivery vehicle scenario in e-Logistics environment, and validate its superior in terms of extensibility and interoperability.