• ETRI Journal
• /
• v.46 no.3
• /
• pp.538-549
• /
• 2024

A conventional codec aims to increase the compression efficiency for transmission and storage while maintaining video quality. However, as the number of platforms using machine vision rapidly increases, a codec that increases the compression efficiency and maintains the accuracy of machine vision tasks must be devised. Hence, the Moving Picture Experts Group created a standardization process for video coding for machines (VCM) to reduce bitrates while maintaining the accuracy of machine vision tasks. In particular, in-loop filters have been developed for improving the subjective quality and machine vision task accuracy. However, the high computational complexity of in-loop filters limits the development of a high-performance VCM architecture. We analyze the effect of an in-loop filter on the VCM performance and propose a suboptimal VCM method based on the selective activation of in-loop filters. The proposed method reduces the computation time for video coding by approximately 5% when using the enhanced compression model and 2% when employing a Versatile Video Coding test model while maintaining the machine vision accuracy and compression efficiency of the VCM architecture.

• Journal of the Korean Society of Industry Convergence
• /
• v.27 no.1
• /
• pp.87-95
• /
• 2024

This study is on modulo scheduling algorithms for multicore processor in machine learning applications. Machine learning algorithms are designed to perform a large amount of operations such as vectors and matrices in order to quickly process large amounts of data stream. To support such large amounts of computations, processor architectures to support applications such as artificial intelligence, neural networks, and machine learning are designed in the form of parallel processing such as multicore. To effectively utilize these multi-core hardware resources, various compiler techniques are being used and studied. In this study, among these compiler techniques, we analyzed the modular scheduler, which is especially important in one core's computation pipeline. This paper looked at and compared the iterative modular scheduler and the swing modular scheduler, which are the most widely used and studied. As a result, both schedulers provided similar performance results, and when measuring register pressure as an indicator, it was confirmed that the swing modulo scheduler provided slightly better performance. In this study, a technique that divides recurrence edge is proposed to improve the minimum initiation interval of the modulo schedulers.

• Journal of Information Processing Systems
• /
• v.14 no.2
• /
• pp.499-509
• /
• 2018

Recently, hybrid transactional memory (HyTM) has gained much interest from researchers because it combines the advantages of hardware transactional memory (HTM) and software transactional memory (STM). To provide the concurrency control of transactions, the existing HyTM-based studies use a bloom filter. However, they fail to overcome the typical false positive errors of a bloom filter. Though the existing studies use a global lock, the efficiency of global lock-based memory allocation is significantly low in multi-core environment. In this paper, we propose an efficient hybrid transactional memory scheme using near-optimal retry computation and sophisticated memory management in order to efficiently process transactions in multi-core environment. First, we propose a near-optimal retry computation algorithm that provides an efficient HTM configuration using machine learning algorithms, according to the characteristic of a given workload. Second, we provide an efficient concurrency control for transactions in different environments by using a sophisticated bloom filter. Third, we propose a memory management scheme being optimized for the CPU cache line, in order to provide a fast transaction processing. Finally, it is shown from our performance evaluation that our HyTM scheme achieves up to 2.5 times better performance by using the Stanford transactional applications for multi-processing (STAMP) benchmarks than the state-of-the-art algorithms.

• Proceedings of the Korean Society of Machine Tool Engineers Conference
• /
• 2004.04a
• /
• pp.235-240
• /
• 2004

The equipment of industrial robot in manufacturing and assembly lines has rapidly increased. In order to achieve high productivity and flexibility, it becomes very important to develop the visual feedback control system with Off-Line Programming System(OLPS). We can save much efforts and time in adjusting robots to newly defined workcells by using OLPS. A proposed visual calibration scheme is based on position-based visual feedback. The calibration program firstly generates predicted images of objects in an assumed end-effector position. The process to generate predicted images consists of projection to screen-coordinates, visible range test and construction of simple silhouette figures. Then camera images acquired are compared with predicted ones for updating position and orientation data. Computation of error is very simple because the scheme is based on perspective projection which can be also expanded to experimental results. Computation time can be extremely reduced because the proposed method does not require the precise calculation of tree-dimensional object data and image Jacobian.

• Proceedings of the KIEE Conference
• /
• 2011.07a
• /
• pp.1210-1211
• /
• 2011

The inherent drawback of iron-core type permanent magnet linear synchronous motor (PMLSM) is detent force that is dependent on several major factors such as PM length, slot clearance, and skewing. To minimize the detent force, this paper proposes a structure optimization using the combination computation of two dimensional (2-D) finite element analysis (FEA) and response surface methodology (RSM). The RSM, that is a collection of the statistical and mathematical techniques, is utilized to predict the global optimal solution based on the FEA calculated results of the detect forces for different combinations of factors. With the help of the combination computation the high capacity iron-core type PMLSM with more than 12000 N propulsion forces only contains less than 3 N detent forces.

• The Journal of the Acoustical Society of Korea
• /
• v.7 no.4
• /
• pp.22-30
• /
• 1988

In this paper, the sound radiation from a clamped circular plate in an infinite baffle is calculated by using the FFT technique. The radiated sound fields are obtained by two-dimensional fast Fourier transform method is the spatial domain instead of a direct numerical evaluation of Rayleigh integral for economy of the computation time. The computation time is consumed at least by 1/200 times of the direct numerical evaluation on the Rayleigh integral in acoustic fields. The FFT method can be applicable to any shaped geometry as well as circular plate. The FFT solution could be very powerful in predicting the near and far fields of complex structures.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.10 no.4
• /
• pp.175-183
• /
• 1985

Dataflow computers exhibit a high degree of parallelism which can not be obtained easily with the conventional von-Neumann architecture. Since many instructions are ready for execution simultaneously, concurrency can be easily achieved by the multiple processors modified the dataflow machine. This paper describes a FFT Butterfly algorithm for dataflow computation and evaluates the performance by the speed up factor of that algorithm through the simulation approach by the time-accelation method.

• Proceedings of the Korean Society of Machine Tool Engineers Conference
• /
• 2003.10a
• /
• pp.213-218
• /
• 2003

The equipment of industrial robot in manufacturing and assembly lines has rapidly increased. In order to achieve high productivity and flexibility, it becomes very important to develop the visual feedback control system with Off-Line Programming System(OLPS). We can save much efforts and time in adjusting robots to newly defined workcells by using OLPS. A proposed visual calibration scheme is based on position-based visual feedback. The calibration program firstly generates predicted images of objects in an assumed end-effector position. The process to generate predicted images consists of projection to screen-coordinates, visible range test and construction of simple silhouette figures. Then camera images acquired are compared with predicted ones for updating position and orientation data. Computation of error is very simple because the scheme is based on perspective projection which can be also expanded to experimental results. Computation time can be extremely reduced because the proposed method does not require the precise calculation of tree-dimensional object data and image Jacobian.

• Proceedings of the Korean Society of Machine Tool Engineers Conference
• /
• 2003.10a
• /
• pp.265-270
• /
• 2003

When designing an engine part the designer has to choose the proper material and to define the dimensions of the part. in the most rudimentary case he has available as guidelines the collective experience from similar applications. The performance of the part is tested on occasion of the trial runs for the full system, and possibly occurring deficiencies are corrected until a satisfactory usable life and safety of operation is achieved. This procedure is time consuming and costly, and the learning effect is minimal. These requirements have to be condensed into characteristic values ameanable to computation. Since testing and computation are complementary we have to look at the currently existing mathematical models for engine bearing performance.

• International Journal of Control, Automation, and Systems
• /
• v.1 no.4
• /
• pp.431-443
• /
• 2003

This paper presents an efficient method to evaluate the performance of an extended stochastic Petri net by simple algebraic operations. The reachability graph is derived from an extended stochastic Petri net, and then converted to a timed stochastic state machine, using a semi-Markov process. The n-th moments of the performance index are derived by algebraic manipulations with each of the n-th moments of transition time and transition probability. For the derivation, three reduction rules are introduced on the transition trajectories in a well-formed regular expression. Efficient computation algorithms are provided to automate the suggested method. The presented method provides a proficient means to derive both the numerical and the symbolic solutions for the performance of an extended stochastic Petri net by simple algebraic manipulations.