• 융합신호처리학회논문지
• /
• 제8권4호
• /
• pp.293-296
• /
• 2007

본 논문에서는 벡터 미디언 값을 계산하기 위한 시스톨릭 어레이 구조의 벡터 미디언 필터 구조를 제안하였다. 컬러영상처리에서 벡터 신호는 빨강, 녹색 파랑의 3개의 요소로 이루어져 있다. 벡터 미디어 필터는 빨강, 녹색 파랑 요소로 이루어진 벡터 신호들 중에서 벡터 신호를 크기 순서대로 나열하였을 때 가운데 값을 갖는 벡터 신호를 구하는 필터로, 컬러 영상처리에서 기본적으로 많이 사용되는 필터이다. 벡터 신호가 N 개가 있을 때, 지금 까지 제안된 구조에서는(3N+1) 클럭이 필요하나, 제안된 구조에서는 (N+2) 클럭이 소요된다. 그리고 기존의 구조에서는 N 개의 입력 벡터 신호는 미디언 필터에 병렬로 입력되어야 하나 제안된 구조에서는 입력 신호는 직렬로 인가된다. FPGA를 사용하여 구현하였다.

• 대한전자공학회:학술대회논문집
• /
• 대한전자공학회 2002년도 하계종합학술대회 논문집(4)
• /
• pp.417-420
• /
• 2002

In this paper, we propose a VLSI architecture for computation of the SVM(Support Vector Machine) that has become established as a powerful technique for solving a variety of classification, regression, and so on. When we compare the proposed systolic arrays with the conventional method, our architecture exhibits a lot of advantages in terms of latency and throughput rate.

• JSTS:Journal of Semiconductor Technology and Science
• /
• 제7권4호
• /
• pp.229-234
• /
• 2007

In this paper, we present an efficient instruction set architecture using vector register file hardware to accelerate operation of general matrix-vector operations in DSP microprocessor. The technique enables in-situ row-access as well as column access to the register files. It can reduce the number of memory access significantly. The technique is especially useful for block-based video signal processing kernels such as FFT/IFFT, DCT/IDCT, and two-dimensional filtering. We have applied the new instruction set architecture to in-loop deblocking filter processing in H.264 decoder. Performance comparisons show that the required load/store operations for the in-loop deblocking filter can be reduced about 42%. The architecture would improve the processing speed, and code density in DSP microprocessor especially for video signal processing substantially.

• 대한전자공학회논문지TC
• /
• 제45권7호
• /
• pp.1-8
• /
• 2008

Low Density Parity Check codes(LDPC)는 최근 우수한 성능으로 통신 분야에서 채널 코딩의 중요한 블록으로 주목받고 있다. 그리하여 Wibro를 포함한 여러 표준에서 LDPC 부호를 채널 코딩으로 채택하고 있다. 이러한 LDPC 부호의 Encoder를 구현하는데 있어서의 약점은 기존의 이진 Matrix Vector Multiplier가 throughput의 감소의 원인이 되는 clock cycle이 많다는 것이다. 본 논문은 표준에서 사용되는 H 행렬이 Circulant Permutation Matrix(CPM)로 정의되어 있다는 점을 이용하여 인코더의 구현에 있어서 기존의 Matrix Vector Multiplier 대신에 cyclic shift register와 exclusive-OR을 사용하는 설계구조를 제안한다. 또한, 제안한 구조를 이용하여 WiBro에 포함되는 다양한 부호율에 적용가능한 인코더를 설계하였다. 제안된 WiBro LDPC의 인코더는 기존보다 적은 clock cycle을 가지므로 높은 throughput에 도달한다.

• 대한전자공학회논문지SD
• /
• 제46권10호
• /
• pp.46-52
• /
• 2009

현재 개발되고 있는 Shader 프로세서는 처리 성능을 높이기 위하여 Multi-Core, Multi-Thread를 채택하고 있다. 또한 Shader 프로세서에서 각 수행 단계별 마다 IP를 따로 구현하지 않고 하나의 Core IP를 다양한 목적으로 사용할 수 있도록 설계하고 있다. 본 논문에서는 이러한 목적에 맞게 Shader-Core를 이용하여 연산이 가능하고, Multi-Core, Multi-Thread 기반에서 픽셀의 병렬처리가 가능하도록 고안된 Vector 기반의 Rasterization알고리즘을 제안한다. 이를 통하여 동일 조건의 기존 알고리즘에 비하여 약 2%의 연산량을 가지면서 각 픽셀이 독립적으로 연산이 가능하도록 하였다.

• JSTS:Journal of Semiconductor Technology and Science
• /
• 제16권4호
• /
• pp.405-414
• /
• 2016

IC-implementations provide high performance for solving the high computational cost of pattern matching but have relative low flexibility for satisfying different applications. In this paper, we report an associative memory architecture for k nearest neighbor (KNN) search, which is one of the most basic algorithms in pattern matching. The designed architecture features reconfigurable vector-component parallelism enabled by programmable switching circuits between vector components, and a dedicated majority vote circuit. In addition, the main time-consuming part of KNN is solved by a clock mapping concept based weighted frequency dividers that drastically reduce the in principle exponential increase of the worst-case search-clock number with the bit width of vector components to only a linear increase. A test chip in 180 nm CMOS technology, which has 32 rows, 8 parallel 8-bit vector-components in each row, consumes altogether in peak 61.4 mW and only 11.9 mW for nearest squared Euclidean distance search (at 45.58 MHz and 1.8 V).

• 전자공학회논문지A
• /
• 제33A권5호
• /
• pp.155-164
• /
• 1996

The architecture of a programmable video codec IC is described that employs multiple vector processors in a single chip. The vector processors operate in parallel and communicate with one another through on-chip shared memories. A single scalar control processor schedules each vector processor independently to achieve real-tiem video coding with special vector instructions. With programmable interconnection buses, the proposed architecture performs multi-processing of tasks and data in video coding. Therefore, it can provide good parallelism as well as good programmability. especially, it can operate multithread video coding, which processes several independent image sequences simultaneously. We explain its scheduling, multithred video coding, and vector processor architectures. We implemented a prototype video codec with a 0.8um CMOS cell-based technology for the multi-standard videophone. This codec can execute video encoding and decoding simultaneously for the QCIF image at a frame rate of 30Hz.

• 대한전기학회논문지:시스템및제어부문D
• /
• 제53권8호
• /
• pp.578-584
• /
• 2004

The CSVM(Current Support Vector Machine) that is a digital architecture performing all phases of recognition process including kernel computing, learning, and recall of SVM(Support Vector Machine) on a chip is proposed. Concurrent operation by parallel architecture of elements generates high speed and throughput. The classification problems of bio data having high dimension are solved fast and easily using the CSVM. Quadratic programming in original SVM learning algorithm is not suitable for hardware implementation, due to its complexity and large memory consumption. Hardware-friendly SVM learning algorithms, kernel adatron and kernel perceptron, are embedded on a chip. Experiments on fixed-point algorithm having quantization error are performed and their results are compared with floating-point algorithm. CSVM implemented on FPGA chip generates fast and accurate results on high dimensional cancer data.

• Management Science and Financial Engineering
• /
• 제17권1호
• /
• pp.19-43
• /
• 2011

Tuning the architecture of SVM (support vector machine) is to build an SVM model of better performance. Two different tuning methods of the grid search and the GA (genetic algorithm) have been addressed in the literature, each of which has its own methodological pros and cons. This paper suggests a combined method for tuning the architecture of SVM models, which employs the GAM (generalized additive models), the grid search, and the GA in sequence. The GAM is used for selecting input variables, and the grid search and the GA are employed for finding optimal parameter values of the SVM models. Applying the method to a bankruptcy prediction problem, we show that SVM model tuned by the proposed method outperforms other SVM models.

• Smart Structures and Systems
• /
• 제15권1호
• /
• pp.57-80
• /
• 2015

Extended Kalman Filter (EKF) has been widely used for structural identification and damage detection. However, conventional EKF approaches require that external excitations are measured. Also, in the conventional EKF, unknown structural parameters are included as an augmented vector in forming the extended state vector. Hence the sizes of extended state vector and state equation are quite large, which suffers from not only large computational effort but also convergence problem for the identification of a large number of unknown parameters. Moreover, such approaches are not suitable for intelligent structural damage detection due to the limited computational power and storage capacities of smart sensors. In this paper, a two-stage and two-step algorithm is proposed for the identification of structural damage as well as unknown external excitations. In stage-one, structural state vector and unknown structural parameters are recursively estimated in a two-step Kalman estimator approach. Then, the unknown external excitations are estimated sequentially by least-squares estimation in stage-two. Therefore, the number of unknown variables to be estimated in each step is reduced and the identification of structural system and unknown excitation are conducted sequentially, which simplify the identification problem and reduces computational efforts significantly. Both numerical simulation examples and lab experimental tests are used to validate the proposed algorithm for the identification of structural damage as well as unknown excitations for structural health monitoring.