• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 1999.06a
• /
• pp.83-88
• /
• 1999

In this paper, a novel gray-scale lossless image coder combining context-based minimum mean squared error (MMSE) prediction and entropy coding is proposed. To obtain context of prediction, this paper first defines directional difference according to sharpness of edge and gradients of localities of image data. Classification of 4 directional differences forms“geometry context”model which characterizes two-dimensional general image behaviors such as directional edge region, smooth region or texture. Based on this context model, adaptive DPCM prediction coefficients are calculated in MMSE sense and the prediction is performed. The MMSE method on context-by-context basis is more in accord with minimum entropy condition, which is one of the major objectives of the predictive coding. In entropy coding stage, context modeling method also gives useful performance. To reduce the statistical redundancy of the residual image, many contexts are preset to take full advantage of conditional probability in entropy coding and merged into small number of context in efficient way for complexity reduction. The proposed lossless coding scheme slightly outperforms the CALIC, which is the state-of-the-art, in compression ratio.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.14 no.12
• /
• pp.4776-4794
• /
• 2020

We propose a rapid adaptive learning framework for streaming object detection, called EER-ASSL. The method combines the expected error reduction (EER) dependent rollback learning and the active semi-supervised learning (ASSL) for a rapid adaptive CNN detector. Most CNN object detectors are built on the assumption of static data distribution. However, images are often noisy and biased, and the data distribution is imbalanced in a real world environment. The proposed method consists of collaborative sampling and EER-ASSL. The EER-ASSL utilizes the active learning (AL) and rollback based semi-supervised learning (SSL). The AL allows us to select more informative and representative samples measuring uncertainty and diversity. The SSL divides the selected streaming image samples into the bins and each bin repeatedly transfers the discriminative knowledge of the EER and CNN models to the next bin until convergence and incorporation with the EER rollback learning algorithm is achieved. The EER models provide a rapid short-term myopic adaptation and the CNN models an incremental long-term performance improvement. EER-ASSL can overcome noisy and biased labels in varying data distribution. Extensive experiments shows that EER-ASSL obtained 70.9 mAP compared to state-of-the-art technology such as Faster RCNN, SSD300, and YOLOv2.

• Journal of the Earthquake Engineering Society of Korea
• /
• v.9 no.1 s.41
• /
• pp.25-32
• /
• 2005

Seismic load is distributed to modes of a structure through the modal participation factor(MPF). The modal participation factor is essential to analyze structural response under earthquake load. MPF of a real structure differs from that of analytical mathematical model due to the error induced from analytical assumptions and during the construction. In this study, an identification method is proposed to calculate the 1st MPF of real structure based on $H^{\infty}$ optimal model reduction. The MPF is obtained from the relationship between observability and controllability matrices realized from system identification and those of a prototype 2-degree state space model. The proposed method is verified thorough numerical examples.

• Proceedings of the KSME Conference
• /
• 2000.11b
• /
• pp.192-198
• /
• 2000

Three dimensional thermal cycle analysis of the plunger is carried out in repeated forming process of the TV glass, which is continued work of two dimensional analysis where an efficient method has been proposed. The plunger undergoes temperature fluctuation during a cycle due to the repeated contact and separation from the glass, which attains a cyclic steady state having same temperature history at every cycle. Straightforward analysis of this problem brings about more than 90 cycles to get reasonable solution. An exponential function fitting method is proposed, which finds exponential function to best approximate temperature values of 3 consecutive cycles, and new cycle is restarted with the fitted value at infinite time. Number of cases are analyzed using the proposed method and compared to the result of straightforward repetition, from which one finds that the method always reaches nearly convergent solution within $9{\sim}12$ cycles, but turns around afterwards without further convergence. Two step use is found most efficient, in which the exponential fitting is carried out fer the first 12 cycles, followed by simple repetition, which shows fast convergence expending only 6 additional cycles to get the accuracy within 2 error. This reduces the computation cycle remarkably from 90 to 18, which is 80% reduction. From the parametric studies, one reveals that the overall thermal behavior of the plunger in terms of cooling parameters and time is similar to that of 2 dimensional analysis.

• Journal of Communications and Networks
• /
• v.18 no.3
• /
• pp.387-396
• /
• 2016

Multiuser communication has been an important research area of underwater acoustic communications and networking. This paper studies the use of adaptive orthogonal frequency-division multiple access (OFDMA) in a downlink scenario, where a central node sends data to multiple distributed nodes simultaneously. In practical implementations, the instantaneous channel state information (CSI) cannot be perfectly known by the central node in time-varying underwater acoustic (UWA) channels, due to the long propagation delays resulting from the low sound speed. In this paper, we explore the CSI feedback for resource allocation. An adaptive power-bit loading algorithm is presented, which assigns subcarriers to different users and allocates power and bits to each subcarrier, aiming to minimize the bit error rate (BER) under power and throughput constraints. Simulation results show considerable performance gains due to adaptive subcarrier allocation and further improvement through power and bit loading, as compared to the non-adaptive interleave subcarrier allocation scheme. In a lake experiment, channel feedback reduction is implemented through subcarrier clustering and uniform quantization. Although the performance gains are not as large as expected, experiment results confirm that adaptive subcarrier allocation schemes based on delayed channel feedback or long term statistics outperform the interleave subcarrier allocation scheme.

• International Area Studies Review
• /
• v.16 no.3
• /
• pp.77-97
• /
• 2012

Estimated coefficients has serious problems including inconsistency, biasness, etc. because many researches about the effect of fiscal decentralization on a country's economic growth use the traditional OLS method. Researches use the data intactly so that so called "spurious regression" phenomenon exists. This causes fundamental fallacy. This research tries unit root test, cointegration test, and then estimates the United States' economic time series by using VECM. The analysis of the effect of the state level-fiscal decentralization on economic growth shows two long term-equilibriums. During short term-dynamic adjustment, fiscal decentralization and economic growth move the same or different directions. In case of prediction GDP increases steeply and then from 2015 gently; and fiscal decentralization index shows a general reduction trend and then decreases slowly. At local level it shows two long term-equilibriums. During short term-dynamic adjustment, fiscal decentralization and economic growth also move the same or different directions. Impulse response analysis shows the very negative effect of fiscal decentralization on economic growth.

• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.21 no.11
• /
• pp.554-560
• /
• 2020

In the manufacturing process of flat panel displays, a high-precision planar motion stage is used to position a specimen. Stages of this type typically use frictionless linear motors and air bearings, and laser interferometers. Real-time dynamic correction of the yaw motion error is very important because the inevitable yaw motion error of the stage means a change in the specimen orientation. Gantry control is generally used to compensate for yaw motion errors. Flexure units that allow rotational motion are applied to the stage to apply this method to a stage using an air-bearing guide. This paper proposes a method to improve the constant speed motion performance of a H-type XY stage equipped with air bearing and flexure units. When applying the gantry control to the stage, including the flexure units, the cause of the mutual ripple generated from the linear motors is analyzed, and adaptive learning control is proposed to compensate for the mutual ripple. A simulation was performed to verify the proposed method. The speed ripple was reduced to approximately the 22 % level. The ripple reduction was verified by simulating the stage state where yaw motion error occurs.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.18 no.1
• /
• pp.79-84
• /
• 2018

This paper relates with the performance improvement of SE-MMA (Signed Error-Multiple Modulus Algorithm) adaptive equalization algorithm that is used for the reduction of the intersymbol interference due to the distortion which occurs in the communication channel for the transmission of 16-QAM nonconstant modulus signal.. In the conventional MMA, the fixed modulus value that is second order statistics of transmitting signal were used, and the SE-MMA was introduced in order to the simplification of the algorithm's arithmetic operation. The SE-MMA have a fast convergence speed than MMA, but it has a problem of degradation of equalization performance in the steady state due to the arithmetic simplification. In this paper, we propose the new algorithm AV-SE-MMA (Adaptively Varying-SE-MMA) that uses the adaptive varying modulus in order to obtain the error signal for updating the adaptive equalizer coefficient, and its equalization performance were confirmed by simulation. In this paper, the performance of SE-MMA and proposed algorithm were compared, and the equalizer output signal constellation, residual isi, MSE and SER in order to confirm the robustness of noise were used as performace index. As a result of performance comparison, the AV-SE-MMA has better performance in output signal constellation, residual isi and MD compared to the SE-MMA, but it was confirmed that the AV-SE-MMA has similar in the SER performance that means the robustness to the noise.

• The Journal of the Acoustical Society of Korea
• /
• v.18 no.6
• /
• pp.49-56
• /
• 1999

In many continuous speech recognition systems based on HMMs, decision tree-based state tying has been used for not only improving the robustness and accuracy of context dependent acoustic modeling but also synthesizing unseen models. To construct the phonetic decision tree, standard method performs one-level pruning using just single Gaussian triphone models. In this paper, two novel approaches, two-level decision tree and multi-mixture decision tree, are proposed to get better performance through more accurate acoustic modeling. Two-level decision tree performs two level pruning for the state tying and the mixture weight tying. Using the second level, the tied states can have different mixture weights based on the similarities in their phonetic contexts. In the second approach, phonetic decision tree continues to be updated with training sequence, mixture splitting and re-estimation. Multi-mixture Gaussian as well as single Gaussian models are used to construct the multi-mixture decision tree. Continuous speech recognition experiment using these approaches on BN-96 and WSJ5k data showed a reduction in word error rate comparing to the standard decision tree based system given similar number of tied states.

• International Journal of Control, Automation, and Systems
• /
• v.4 no.2
• /
• pp.204-216
• /
• 2006

Latest advances in hardware technology and state of the art of mobile robot and artificial intelligence research can be employed to develop autonomous and distributed monitoring systems. And mobile service robot requires the perception of its present position to coexist with humans and support humans effectively in populated environments. To realize these abilities, robot needs to keep track of relevant changes in the environment. This paper proposes a localization of mobile robot using the images by distributed intelligent networked devices (DINDs) in intelligent space (ISpace) is used in order to achieve these goals. This scheme combines data from the observed position using dead-reckoning sensors and the estimated position using images of moving object, such as those of a walking human, used to determine the moving location of a mobile robot. The moving object is assumed to be a point-object and projected onto an image plane to form a geometrical constraint equation that provides position data of the object based on the kinematics of the intelligent space. Using the a priori known path of a moving object and a perspective camera model, the geometric constraint equations that represent the relation between image frame coordinates of a moving object and the estimated position of the robot are derived. The proposed method utilizes the error between the observed and estimated image coordinates to localize the mobile robot, and the Kalman filtering scheme is used to estimate the location of moving robot. The proposed approach is applied for a mobile robot in ISpace to show the reduction of uncertainty in the determining of the location of the mobile robot. Its performance is verified by computer simulation and experiment.