• Journal of the Korea Computer Graphics Society
• /
• v.30 no.3
• /
• pp.99-107
• /
• 2024

While the demand for extrapolating video content horizontally or vertically is increasing, even the most advanced techniques cannot successfully extrapolate all videos. Therefore, it is important to determine if a given video can be well extrapolated before attempting the actual extrapolation. This can help avoid wasting computing resources. This paper proposes a video classifier that can identify if a video is suitable for horizontal extrapolation. The classifier utilizes optical flow and an adaptive Gaussian blur network, which can be applied to flow-based video extrapolation methods. The labeling for training was rigorously conducted through user tests and quantitative evaluations. As a result of learning from this labeled dataset, a network was developed to determine the extrapolation capability of a given video. The proposed classifier achieved much more accurate classification performance than methods that simply use the original video or fixed blur alone by effectively capturing the characteristics of the video through optical flow and adaptive Gaussian blur network. This classifier can be utilized in various fields in conjunction with automatic video extrapolation techniques for immersive viewing experiences.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.19 no.2
• /
• pp.211-224
• /
• 1994

In this paper, a multi-level adaptive intraframe method has been proposed to separate the luminance and chominance components in NTSC composite signal. The control signals are generated by detecting the vertical correlation and transition in the horizontal and diagonal directions. The chrominance component is adaptively processed through vertical and horizontal filters according to the control signals and the luminance component is processed by subtracting the chrominance component from the composite video signal. The several filters have been used at the sampling rate of four times the color subcarrier frequency and computer simulation and SVP(Serial Video Processing) system have been introduced to compare the performance of the conventional methods and that of proposed one.

• The KIPS Transactions:PartC
• /
• v.12C no.6 s.102
• /
• pp.901-906
• /
• 2005

In this paper, we propose a novel MHCOC (Mapped High Capacity Orthogonal Code) SS(Spread Spectrum) technique that reduces high PAPR (Peak power to Average Power Ratio) of HCOC SS system which was proposed to support high data rate transmission, and we compare to the conventional modulation technique such as MQAM SS that can transmit the same number of symbols at the same time. Moreover, we study on adaptation of this system to satisfy QoS (Quality of Service) that services the proper data rate according to the propagation channel quality information. We perform computer simulation to verify the performance of the proposed system and analyze its availability.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2008.11a
• /
• pp.1236-1239
• /
• 2008

적응형 대역폭 할당 방법에서는 가변 비트율의 비디오 데이터에 대해 스무딩 알고리즘을 사용하여 전송 계획을 세우고, 이 계획에 따라 데이터를 네트워크 트래픽을 고려하여 전송한다. 그러나, 새로운 구간에서 높은 전송률이 요구될 수 있는 개선된 e-PCRTT 알고리즘[4]를 이 방법에서의 전송 계획으로 그대로 사용하려는 경우 가용 전송률이 낮은 경우에는 폐기되는 프레임의 양이 많아질 수 있다. 이를 개선한 알고리즘[5]에서는 언더플로우 경계선과 오버플로우 경계선의 중간에서 전송률이 변화되지만, 전송률의 증가가 필요한 경우에는 전송률 증가 양을 최소로 한다. 본 논문에서는, 개선된 e-PCRTT 알고리즘[4]와 이를 다시 개선한 알고리즘[5]의 성능을 평가하기 위하여 E.T.90 비디오 데이터를 사용하여, 첨두 전송률, 평균 전송률 증가양, 폐기되는 프레임 양을 비교하였다.

• Proceedings of the Korean Institute of Intelligent Systems Conference
• /
• 2004.04a
• /
• pp.261-264
• /
• 2004

컴퓨터 전략 시뮬레이션 게임 설계에서는 Master ADU(Artificial Decision Unit)의 전략 수립을 위한 방법으로 다양한 기법들이 연구되고 있다. 특히 한정된 자원 하에서 게임을 사실적이고 지적인 기능을 구현하기 위해 치팅(Cheating)을 활용하거나 간단한 인공지능 기법이 적용되고 있다. 하지만 이 기법들은 사용자 적응성 및 전략 수립의 단순성을 야기하는 단점을 가지고 있다. 본 연구에서는 전략 시뮬레이션 게임의 전략 수립 에이전트인 Master ADU(Artificial Decision Unit)를 위하여 셀룰라 오토마타의 초기 규칙 생성에 유전자 알고리즘의 교배 및 돌연변이, 적합도 평가를 거친 유전자 형을 적용한 혼합형 전략 수립 기법을 제안한다 이 기법은 ADU가 적합한 유전자 형을 생산 및 선택하여 사용자에 대해 적극적으로 학습할 수 있었다.

• Journal of the Korea Computer Industry Society
• /
• v.3 no.9
• /
• pp.1251-1260
• /
• 2002

Real-time application programs have synchronization constraints which need to be met between media-data. Synchronization method represents feedback method including virtual client-side buffer. This buffer is used in buffer level method. It is client-leading synchronization that is absorbing variable transmission delay time and that is synchronizing by feedback control. It is the important factor for playback rate and QoS if the buffer level is normal or not. To solve the problems, we can control the start of transmission in multimedia server by appling filtering, control and network evaluation function. Synchronization method is processing for smooth presentation without cut-off while media is playing out. When audio frame which is master media is in high threshold buffer level we decrease play out time gradually, otherwise we increase it slowly.

• Journal of the Korea Society of Computer and Information
• /
• v.16 no.7
• /
• pp.105-115
• /
• 2011

In Vision-based Intelligent Transportation System(ITS) the segmentation of road region is a very basic functionality. Accordingly, in this paper, we propose a region segmentation method using adaptive pattern extraction technique to segment road regions and sky regions from original images. The proposed method consists of three steps; firstly we perform the initial segmentation using Mean Shift algorithm, the second step is the candidate region selection based on a static-pattern matching technique and the third is the region growing step based on a dynamic-pattern matching technique. The proposed method is able to get more reliable results than the classic region segmentation methods which are based on existing split and merge strategy. The reason for the better results is because we use adaptive patterns extracted from neighboring regions of the current segmented regions to measure the region homogeneity. To evaluate advantages of the proposed method, we compared our method with the classical pattern matching method using static-patterns. In the experiments, the proposed method was proved that the better performance of 8.12% was achieved when we used adaptive patterns instead of static-patterns. We expect that the proposed method can segment road and sky areas in the various road condition in stable, and take an important role in the vision-based ITS applications.

• The KIPS Transactions:PartA
• /
• v.14A no.6
• /
• pp.383-390
• /
• 2007

This paper describes the e-learning system for science and engineering mathematics using computer algebra system and Bayesian inference network. The best feature of this system is using one of the most recent mathematical dynamic web content authoring model which is called client independent dynamic web content authoring model and using the Bayesian inference network for diagnosing student's learning. The authoring module using computer algebra system provides teacher-user with easy way to make dynamic mathematical web contents. The diagnosis module using Bayesian inference network helps students know the weaker parts of their learning, in this way our system determines appropriate next learning sequences in order to provide supplementary learning feedback.

• Proceedings of the Korean Society of Computer Information Conference
• /
• 2024.01a
• /
• pp.51-54
• /
• 2024

본 논문에서는 온 디바이스 국방 AI를 위한 효율적인 학습 방법을 제안한다. 제안하는 방법은 모델 전체를 재학습하는 대신 필요한 부분만 세밀하게 조정하여 계산 비용과 시간을 대폭 줄이는 PEFT 기법의 LoRa를 적용하였다. LoRa는 기존의 신경망 가중치를 직접 수정하지 않고 추가적인 낮은 랭크의 매트릭스를 학습하는 방식으로 기존 모델의 구조를 크게 변경하지 않으면서도, 효율적으로 새로운 작업에 적응할 수 있다. 또한 학습 파라미터 및 연산 입출력에 데이터에 대하여 32비트의 부동소수점(FP32) 대신 부동소수점(FP16, FP8) 또는 정수형(INT8)을 활용하는 경량화 기법인 양자화도 적용하였다. 적용 결과 학습시 요구되는 GPU의 사용량이 32GB에서 5.7GB로 82.19% 감소함을 확인하였다. 동일한 조건에서 동일한 데이터로 모델의 성능을 평가한 결과 동일 학습 횟수에선 LoRa와 양자화가 적용된 모델의 오류가 기본 모델보다 53.34% 증가함을 확인하였다. 모델 성능의 감소를 줄이기 위해서는 학습 횟수를 더 증가시킨 결과 오류 증가율이 29.29%로 동일 학습 횟수보다 더 줄어듬을 확인하였다.

• Journal of the Korea Society of Computer and Information
• /
• v.6 no.4
• /
• pp.94-99
• /
• 2001

Recently a lot of informations is interchanged among many users through the network such as Internet. In this situation. multimedia data transfer has some problems. that are jitter, the difference of delay in arriving packets, and loss of data due to the various delay between sender and receiver. The arriving data packets are not synchronized because of those problems. Especially, an efficient method is needed to revise the sudden large changes of the delay, called spike-like delay. which is occured in explosive growth of networking. We propose efficient synchronization algorithm which controls synchronization interval and adjusts to sudden changes in networks. The algorithm, a receiver-driven adaptive synchronization method, is to synchronize the arriving data packets against spike-like delay and decrease the loss of them. In addition, another method is proposed in this paper. The method uses the probabilistic features of delay distribution appearing in general network. It is evaluated for the loss of data packets and the delay times.