• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2019.11a
• /
• pp.119-121
• /
• 2019

부가적 백색 잡음 모델(additive white Gaussian noise, AWGN에서 학습된 깊은 신경만 (deep neural networks)을 이용한 잡음 제거기는 제거하려는 잡음이 AWGN인 경우에는 뛰어난 성능을 보이지만 실제 카메라 잡음에 대해서 잡음 제거를 시도하였을 때는 성능이 크게 저하된다. 본 논문은 U-Net 구조의 깊은 인공신경망 모델에 residual block을 결합함으로서 실제 카메라 영상에서 기존 알고리즘보다 뛰어난 성능을 지니는 신경망을 제안하다. 제안한 방법을 통해 Darmstadt Noise Dataset에서 PSNR과 SSIM 모두 CBDNet 대비 향상됨을 확인하였다.

• Proceedings of the Korea Electromagnetic Engineering Society Conference
• /
• 2000.11a
• /
• pp.259-263
• /
• 2000

21기 정보화사회를 맞이하여 전과서비스 사용의 증가로 인공전파잡음이 계속 증가하는 추세이므로 잡음원과 전파환경 분포를 파악하기 위하여 각 지역의 전파잡음레벨을 조사하고 있다. 그 중 년도별 전파잡음레벨의 변화와 시간에 따른 잡음레벨 변화를 분석해보니 주파수 대역에 따라 특징적인 변화를 보여주고 있다. 따라서 앞으로 주파수 스펙트럼분포와 잡음원과의 관계를 분석하여 전파환경 보호에 관한 대책을 세우고 원활한 전파서비스가 제공될 수 있도록 할 것이다.

• Journal of the Korean Society of Radiology
• /
• v.12 no.2
• /
• pp.193-199
• /
• 2018

In this research, 15 patients were diagnosed with 1.5T and 3.0T MRI instruments (Philips, Medical System, Achieva) to minize Ferromagnetic artifact and find the optimized Tesla. Based on the theory that the 3.0T, when compared to 1.5T, show relatively high signal-to-ratio(SNR), Scan time can be shortened or adjust the image resolution. However, when using the 3.0T MRI instruments, various artifact due to the magnetic field difference can degrade the diagnostic information. For the analysis condition, area of interest is set at the background of the T1, T2 sagittal image followed by evaluation of L3, L4, L5 SNR, length of 3 parts with Ferromagnetic artifact, and Histogram. The validity evaluation was performed by using the independent t test. As a result, for the SNR evaluation, mere difference in value was observed for L3 between 1.5T and 3.0T, while big differences were observed for both L4, and L5(p<0.05). Shorter length was observed for the 1.5T when observing 3 parts with Ferromagnetic artifact, thus we can conclude that 3.0T can provide more information on about peripheral tissue diagnostic information(p<0.05). Finally, 1.5T showed higher counts values for the Histogram evaluation(p<0.05). As a result, when we have compared the 1.5T and 3.0T with SNR, length of Ferromagnetic artifact, Histogram, we believe that using a Low Tesla for Spine MRI test can achieve the optimal image information for patients with disk operation like PLIF, etc. in the past.

• Geophysics and Geophysical Exploration
• /
• v.25 no.2
• /
• pp.71-84
• /
• 2022

When acquiring seismic data, various types of simultaneously recorded seismic noise hinder accurate interpretation. Therefore, it is essential to attenuate this noise during the processing of seismic data and research on seismic noise attenuation. For this purpose, machine learning is extensively used. This study attempts to attenuate noise in prestack seismic data using unsupervised machine learning. Three unsupervised machine learning models, N2NUNET, PATCHUNET, and DDUL, are trained and applied to synthetic and field prestack seismic data to attenuate the noise and leave clean seismic data. The results are qualitatively and quantitatively analyzed and demonstrated that all three unsupervised learning models succeeded in removing seismic noise from both synthetic and field data. Of the three, the N2NUNET model performed the worst, and the PATCHUNET and DDUL models produced almost identical results, although the DDUL model performed slightly better.

• The Journal of Korean Institute of Electromagnetic Engineering and Science
• /
• v.21 no.11
• /
• pp.1295-1303
• /
• 2010

In this paper, an effective method has been proposed for selection of white Gaussian noise(WGN) band for radio background noise measurement system under broad band noise environment. MMN which comes from industrial devices and equipment mostly happens in the shape of broad band noise mostly like impulsive noise and this is the main reason for increasing level in the present radio noise measurements. The existing method based on singular value decomposition has weak point that it cannot give good performance for the broad band signal because it uses signal's white property. The proposed method overcomes such a weakness of singular value decomposition based method by using signal's Gaussian property based method in parallel. Moreover, this proposed method hires a modelling based method which uses parameter estimation algorithm like maximum likelihood estimation(MLE) and gives more accurate result than the method using amplitude probability distribution(APD) graph. Experiment results under the natural environment has done to verify feasibility of the proposed method.

• Journal of Advanced Navigation Technology
• /
• v.24 no.1
• /
• pp.47-52
• /
• 2020

Direction of arrival (DoA) estimation is a scheme of estimating the directions of targets by analyzing signals generated or reflected from the targets and is used in various fields. Artificial neural networks (ANN) is a field of machine learning that mimics the neural network of living organisms. They show good performance in pattern recognition. Although researches has been using ANNs to estimate the DoAs, there are limitationsin dealing with variations of the signal-to-noise ratio (SNR) of the target signals. In this paper, we propose a three-stage ANN algorithm for DoA estimation. The proposed algorithm can minimize the performance reduction by applying the model trained in a single SNR environment to various environments through a 'noise reduction process'. Furthermore, the scheme reduces the difficulty in learning and maintains efficiency in estimation, by employing a process of DoA shift. We compare the performance of the proposed algorithm with Cramer-Rao bound (CRB) and the performances of existing subspace-based algorithms and show that the proposed scheme exhibits better performance than other schemes in some severe environments such as low SNR environments or situations in which targets are located very close to each other.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.43 no.5 s.311
• /
• pp.79-87
• /
• 2006

In this study, we evaluated the performance of a companding strategy as a preprocessing for speech enhancement and noise reduction. The proposed algorithm is based on two tone suppression that is human's hearing characteristics. This algorithm enhances spectral peak of speech signal and reduces background noise, however it has tradeoff characteristics between speech distortion and noise reduction due to limited channel number and nonlinear block. Therefore, we designed two different companding structures that have relative characteristics of noise reduction and speech distortion and found suitable companding structures by difference of individual speech perception ability in noise environment. Thus we proposed speech perception enhancement of cochlear implant user in noise environment with low SNR. The performance of the proposed algorithm was evaluated through 5 normal hearing listeners using noise band simulation. Improvement of speech perception was observed for all subjects and each subject preferred the different type of companding structure.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2022.11a
• /
• pp.550-552
• /
• 2022

최근 사물인터넷(IoT) 기기가 활성화됨에 따라 웨어러블 장치 환경에서 장기간 모니터링 및 수집이 가능해짐에 따라 생체 신호 처리 및 ECG 분석 연구가 활성화되고 있다. 그러나, ECG 데이터는 부정맥 비트의 불규칙적인 발생으로 인한 클래스 불균형 문제와 근육의 떨림 및 신호의 미약등과 같은 잡음으로 인해 낮은 신호 품질이 발생할 수 있으며 훈련용 공개데이터 세트가 작다는 특징을 갖는다. 이 논문에서는 ECG 1D 신호를 2D 스펙트로그램 이미지로 변환하여 잡음의 영향을 최소화하고 전이학습과 메타학습의 장점을 결합하여 클래스 불균형 문제와 소수의 데이터에서도 빠른 학습이 가능하다는 특징을 갖는다. 따라서, 이 논문에서는 ECG 스펙트럼 이미지를 사용하여 2D-CNN 메타-전이 학습 기반 부정맥 분류 기법을 제안한다.

• The Journal of Korean Institute of Electromagnetic Engineering and Science
• /
• v.9 no.4
• /
• pp.462-472
• /
• 1998

In this paper, considering the Rayleigh fading, the lognormal shadowing, and the man-made noise which is occurred in urban area randomly, the mobile radio channel and the radio propagations are analyzed. The system affected by the noise is compared to other modelings. The fading, shadowing, and background noise are wholly considered to evaluate the mobile radio propagation effectively. For N=0.000001, the outage probability in the absence of noise is $5.264{times}10^{-6}$, in the fading only $3.1796{times}10^{-4}$, and in the presence of noise $6.0{\times}10^{-3}$. The analysis with the presence of noise is very important for the performance evaluation of mobile radio system.

• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.11 no.3
• /
• pp.965-970
• /
• 2010

In this paper, a iterative image restoration method is proposed to restore for noisy-blurred images. In conventional method, regularization is usually applied to all over the without considering the local characteristics of image. As a result, ringing artifacts appear in edge regions and the noise amplification is introduced in flat regions. To solvethis problem we proposed an adaptive regularization iterative restoration using directional regularization operator considering edges in four directions and the regularization operator with no direction for flat regions. We verified that the proposed methods showed better results in the suppression of the noise amplification in flat regions, and introduced less ringing artifacts in edge regions. As a result it showed visually better image and improved better ISNR further than the conventional methods.