• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.22 no.5
• /
• pp.561-566
• /
• 2021

The optimum kernel size for error-distribution estimation with given error samples cannot be used in the weight adjustment of minimum Euclidean distance between error distributions (MED) algorithms. In this paper, a new adaptive kernel estimation method for convergence enhancement of MED algorithms is proposed. The proposed method uses the average rate of change in error power with respect to a small interval of the kernel width for weight adjustment of the MED learning algorithm. The proposed kernel adjustment method is applied to experiments in communication channel compensation, and performance improvement is demonstrated. Unlike the conventional method yielding a very small kernel calculated through optimum estimation of error distribution, the proposed method converges to an appropriate kernel size for weight adjustment of the MED algorithm. The experimental results confirm that the proposed kernel estimation method for MED can be considered a method that can solve the sensitivity problem from choosing an appropriate kernel size for the MED algorithm.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.9 no.5
• /
• pp.549-554
• /
• 2014

In the paper, we propose a new Signed LMS(Least Mean Square) algorithm for ICS(Interference Cancellation System). The proposed Signed LMS algorithm improved performances by adjusting step size values. At the convergence of 1000 iteration state, the MSE(Mean Square Error) performance of the proposed Signed LMS algorithm with step size of 0.067 is about 3 ~ 18 dB better than the conventional LMS, CMA algorithm. And the proposed Signed LMS algorithm requires 500 ~ 4000 less iterations than the and LMS and CMA algorithms at MSE of -25dB.

• Journal of IKEEE
• /
• v.6 no.2 s.11
• /
• pp.111-118
• /
• 2002

In this paper, we present a variable block method to reduce additive computational requirements in determining step-size of variable step-size LMS (VS-LMS) algorithms. The block length is inversely proportional to the changing of step-size in VS-LMS algorithm. The technique reduces computational requirements of the conventional VS-LMS algorithms without a degradation of performance in convergence rate and steady state error. And a method for deriving initial step-size, when the input is zero mean, white Gaussian sequence, is proposed. For demonstrating the good performances of the proposed method, simulation results are compared with the conventional variable step-size algorithms in convergence speed and computational requirements.

• Journal of the Korean Society of Clothing and Textiles
• /
• v.26 no.3_4
• /
• pp.390-401
• /
• 2002

This study was to evaluate the fitness and the suitability of size specification of the ready made jacket for women by analysing pattern size and space between skin and clothing using 3D scanner. The results were as follows: 1. Pattern B had the lowest score and the feeling of wearing was significantly different among the given patterns even though all jacket size specification were the same. 2. Ease amount was different between each brand even though the jacket size specification was the same due to the different pattern grading rules. And increasing grading amounts were bigger in horizontal direction rather than in vertical direction. 3. We could obtain accurate a 3 dimensional figure, using 3D scanner which was very useful and more accurate than 2 dimensional data using photography method. 4. Analyzing the average space between skin and clothing of each pattern, there was no significant difference in the average space between skin and clothing among all patterns except waist part of B88 size. And analyzing the average space between skin and clothing of each measured body parts by each size, there existed a significant difference in the interscye, abdomen and hip parts.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.15 no.6
• /
• pp.987-994
• /
• 2020

In this paper, we studied the effect of kernel size of CNN layer on performance in acoustic noise attenuators. This system uses a deep learning algorithm using a neural network adaptive prediction filter instead of using the existing adaptive filter. Speech is estimated from a single input speech signal containing noise using a 100-neuron, 16-filter CNN filter and an error back propagation algorithm. This is to use the quasi-periodic property in the voiced sound section of the voice signal. In this study, a simulation program using Tensorflow and Keras libraries was written and a simulation was performed to verify the performance of the noise attenuator for the kernel size. As a result of the simulation, when the kernel size is about 16, the MSE and MAE values are the smallest, and when the size is smaller or larger than 16, the MSE and MAE values increase. It can be seen that in the case of an speech signal, the features can be best captured when the kernel size is about 16.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.26 no.3
• /
• pp.443-448
• /
• 2022

This paper presents a subband affine projection algorithm with variable step size to improve convergence performance in adaptive filtering applications with long adaptive filters and highly correlated input signals. The proposed algorithm can obtain fast convergence speed and small steady-state error by using different step sizes for each adaptive sub-filter in the subband structure to which polyphase decomposition and noble identity are applied. The step size derived to minimize the mean square error of the adaptive filter at each update time shows better convergence performance than the existing algorithm using a variable step size. In order to confirm the convergence performance of the proposed algorithm, which is superior to the existing algorithm, computer simulations are performed for mean square deviation(MSD) for AR(1) and AR(2) colored input signals considering the system identification model.

• Proceedings of the Korean Information Science Society Conference
• /
• 2011.06b
• /
• pp.338-341
• /
• 2011

프로그램 코드 사이즈는 내장형시스템 구성에 있어서 고려해야 할 핵심 요소중의 하나이다. 프로그램 사이즈는 해당 시스템의 메모리 크기, 전력소모, 성능, 가격 등에 영향을 미치기 때문이다. 프로그램 코드 사이즈를 최적화하기 위하여 활용할 수 있는 시스템 자원 중에서 효과적인 것 중 하나가 복수 로드/스토어 명령어(Multiple Load/Store Instruction, MLS)이다. MLS 명령어는 하나의 명령어로 하나이상의 메모리 값을 레지스터로 블록 전송 (block transfer)하는 것이 가능하기 때문이다. 본 연구에서는 MLS명령어를 기존보다 효과적으로 생성함으로써 코드 크기를 감소시키는 최적화 기법에 대해 논의한다. 실험을 통하여 Mediabench와 DSPStone 벤치마크에서 본 연구에서 제안하는 기법을 통하여 평균 메모리 접근 코드사이즈가 10.3% 감소하였다.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.14 no.5
• /
• pp.805-810
• /
• 2019

Recently, a majority of 4G mobile telecommunication manufacturers prefer repeaters with good adaptability. In this paper, we propose a new LMF(: Least Means Fourth) algorithm for LTE(: Long Term Evolution) RF(: Radio Frequency) Repeater. The proposed algorithm is a modification of the LMF, which appropriately adjusts the step size and improves performance according to the Sign function. The steady state MSE(: Mean Square Error) performance of the proposed LMF algorithm with step size of 0.009 is low level at about -25dB, and the proposed LMF algorithm requires 500 less iterations than the conventional algorithms at MSE of -25dB.

• Journal of the Korean Society of Radiology
• /
• v.15 no.4
• /
• pp.547-554
• /
• 2021

In order to quickly and accurately diagnose pneumonia on a chest X-ray image, different batch sizes of 4, 8, 16, and 32 were applied to the same Xception deep learning model, and modeling was performed 3 times, respectively. As a result of the performance evaluation of deep learning modeling, in the case of modeling to which batch size 32 was applied, the results of accuracy, loss function value, mean square error, and learning time per epoch showed the best results. And in the accuracy evaluation of the Test Metric, the modeling applied with batch size 8 showed the best results, and the precision evaluation showed excellent results in all batch sizes. In the recall evaluation, modeling applied with batch size 16 showed the best results, and for F1-score, modeling applied with batch size 16 showed the best results. And the AUC score evaluation was the same for all batch sizes. Based on these results, deep learning modeling with batch size 32 showed high accuracy, stable artificial neural network learning, and excellent speed. It is thought that accurate and rapid lesion detection will be possible if a batch size of 32 is applied in an automatic diagnosis study for feature extraction and classification of pneumonia in chest X-ray images using deep learning in the future.

• Journal of the Korean Society of Clothing and Textiles
• /
• v.25 no.10
• /
• pp.1707-1718
• /
• 2001

본 연구 목적은 3차원 인체 스캐너를 이용하여 여자 기성복 재킷의 여유량을 비교, 분석하는 것으로, 2사이즈 7브랜드의 재킷의 공극량을 계측하여 분석하였다. 첫째, 재킷 단면둘레 분석 결과, B85(품), B8(허리)를 제외하고 브랜드간에 유의한 차이를 나타내지 않아, 전반적으로, 브랜드간 제품치수에는 차이가 없는 것으로 나타났다. 둘째, 인체와 재킷의 단면둘레 분석 결과, 재킷의 배둘레를 제외한 모든 항목에서 유의한 차이가 나타나 피험자, 재킷 모두 사이즈에 따라 유의적인 차이 가 있음을 알 수 있다. 셋째, 기본사이즈 B85에서는 허리를 제외하고는 패턴 F가 가장 여유량이 많은 것으로 나타났으나, B88의 경우, 부위별로 각기 다른 패턴에서 여유량이 가장 많은 것으로 나타나, 각 부분마다 브랜드별로 그레이딩 룰이 다름을 알 수 있다. 넷째, 착의 단면은 인체와 의복간의 여유량 분포를 명백히 보여주며, 어깨, 가슴, 엉덩이처럼 몸에 밀착되는 부위는 다른 부위에 비해 패턴간, 각도별 변이가 적은 것으로 나타났다. 품, 허리, 배에서는 옆보다는 앞, 뒤로, 가슴에서는 앞뒤 좌우의 30$^{\circ}$방향, 엉덩이의 경우, 옆, 뒤보다는 앞쪽에 여유량이 집중되어 있는 것으로 나타났다. 다섯째, 브랜드별 평균공극길이에 대한 분산분석 결과, 전반적으로 패턴 F가 가장 공극량이 많고, 패턴 D가 작은 것으로 나타났다. 여섯째, 사이즈별 평균공극길이에 대한 t-검정 결과, 품과 배 부분에서, B88이 B85보다 공극량이 적은 것으로 나타나, 기준부위인 가슴, 허리, 엉덩이 부분뿐만 아니라 품, 배둘레의 치수에 대응할 수 있도록 그레이딩 룰 값을 산정하여야 함을 알 수 있다.