• 한국음향학회지
• /
• 제42권5호
• /
• pp.460-468
• /
• 2023

고관절 이형성증(Developmental Dysplasia of Hip, DDH)은 영유아 성장기에 흔히 발생하는 병리학적 상태로, 영유아의 성장을 방해하고 잠재적인 합병증을 유발하는 원인 중 하나이며 이를 조기에 발견하고 치료하는 것은 매우 중요하다. 기존의 DDH 진단 방법으로는 촉진법과 X-ray 또는 초음파 영상 기반 고관절에서의 특징점 검출을 이용한 진단 방법이 있지만 특징점 검출 시 객관성과 생산성에 제한점이 존재한다. 본 연구에서는 X-ray 및 초음파 영상을 이용한 딥러닝 모델 기반 특징점 검출 방법을 제시하고, 다양한 딥러닝 모델을 이용하여 특징점 검출의 성능을 비교 분석하였다. 또한, 부족한 의료 데이터를 보완하는 방법인 다양한 데이터 증강 기법을 제시하고 비교 평가하였다. 본 연구에서는 Residual Network 152(ResNet152) 및 Simple & Complex augmentation 기법을 적용하였을 때 가장 높은 특징점 검출 성능을 보여주었으며, X-ray 영상에서 평균 Object Keypoint Similarity(OKS)가 약 95.33 %, 초음파 영상에서는 약 81.21 %로 각각 측정되었다. 이러한 결과는 고관절 초음파 및 X-ray 영상에서 딥러닝 모델을 적용함으로써 DDH 진단 시 특징점 검출에 관한 객관성과 생산성을 향상시킬 수 있음을 보여준다.

• 한국음향학회지
• /
• 제21권6호
• /
• pp.522-530
• /
• 2002

본 논문에서는 선형 예측 후에 얻어지는 잔차 신호 (residual signal)를 신경 회로망에 바탕을 둔 비선형 예측기로 예측하는 방법을 제안하였다. 신경 회로망을 이용한 예측 방법의 타당성을 입증하기 위해, 먼저 선형 장구간 예측기와 신경 회로망이 도입된 비선형 장구간 예측기의 성능을 서로 비교하였다. 그리고 비선형 예측 후의 잔차 신호를 양자화 하는 과정에서 발생하는 양자화 오차의 영향에 대해 분석하였다. 제안된 신경망 예측기는 예측 오차뿐만 아니라 양자화의 영향을 함께 고려하였으며, 양자화오차에 대한강인성을 갖게 하기 위하여 쿤-터커 (Kuhn-Tucker) 부등식 조건을 만족하는 제한조건 역전파 알고리즘을 새로이 제안하였다. 실험 결과, 제안된 신경망 예측기는 제한조건을 갖는 학습 알고리즘을 사용했음에도 불구하고, 예측 이득이 크게 뒤떨어지지 않는 성능을 나타내었다.

• 전자공학회논문지SC
• /
• 제40권6호
• /
• pp.1-7
• /
• 2003

본 논문은 비선형 시스템에서 시스템의 출력이 주기적인 특징을 가지는 궤적을 따라가도록 하는 슬라이딩 모드 반복 제어기를 소개한다. 본 논문에서 제안하는 제어기는 전체 시스템을 안정화시키며 출력오차를 어떤 범위 안으로 지수적으로 수렴시키는 슬라이딩 모드 제어기와 추적 오차의 수렴을 위해서 사용되는 반복 제어기로 구성되어 있다. 본 논문에서는 제안하는 슬라이딩 모드 제어기는 기존의 방법과는 다르게 제어입력의 크기가 추적오차의 크기에 비례하게 되어 있어서 시스템의 차수를 올리지 않고 정상상태에서의 채터링(chattering) 문제를 크게 개선하는 특징을 가지고 있다.

• E2M - 전기 전자와 첨단 소재
• /
• 제11권11호
• /
• pp.1-8
• /
• 1998

This paper presents an arificial neuro-fuzzy technique based prtial discharge (PD) pattern classifier to power system application. This may require a complicated analysis method employ -ing an experts system due to very complex progressing discharge form under exter-nal stress. After referring briefly to the developments of artificical neural network based PD measurements, the paper outlines how the introduction of new emerging technology has resulted in the design of a number of PD diagnostic systems for practical applicaton of residual lifetime prediction. The appropriate PD data base structure and selection of learning data size of PD pattern based on fractal dimentsional and 3-D PD-normalization, extraction of relevant characteristic fea-ture of PD recognition are discussed. Some practical aspects encountered with unknown stress in the neuro-fuzzy techniques based real time PD recognition are also addressed.

• 한국방송∙미디어공학회:학술대회논문집
• /
• 한국방송∙미디어공학회 2021년도 추계학술대회
• /
• pp.91-94
• /
• 2021

As deep learning technologies becoming developed, realistic fake videos synthesized by deep learning models called "Deepfake" videos became even more difficult to distinguish from original videos. As fake news or Deepfake blackmailing are causing confusion and serious problems, this paper suggests a novel model detecting Deepfake videos. We chose Residual Convolutional Neural Network (Resnet50) as an extraction model and Long Short-Term Memory (LSTM) which is a form of Recurrent Neural Network (RNN) as a classification model. We adopted cosine similarity with hinge loss to train our extraction model in embedding the features of Deepfake and original video. The result in this paper demonstrates that temporal features in the videos are essential for detecting Deepfake videos.

• 한국수자원학회:학술대회논문집
• /
• 한국수자원학회 2020년도 학술발표회
• /
• pp.365-365
• /
• 2020

최근 기후변화로 인한 강우, 온도, 유량과 같은 수문학적 요소의 불확실성 증가와 더불어 산업화, 도시화로 인한 물 수요가 커짐에 따라 물부족 발생 위험이 증가하고 있다. 이에 따라, 안정적인 물 공급을 위한 하천유량과 취수량의 균형을 목적으로 하는 취수량의 예측 및 모의에 대한 중요성이 강조되고 있다. 본 연구에서는 과거 하천 취수량 자료로부터 미래 취수량을 예측하기 위해 딥러링 기법 중 하나인 순환신경망(LSTM) 모형과 시계열분해법을 결합하여 취수량 예측 모형을 개발하였다. 시계열분해법을 통해 자료의 경향성과 계절적 변동성 등 다양한 스케일의 시계열을 분해하여 전처리를 수행하였으며 불확실성을 의미하는 잔차(residual)에 LSTM을 적용하여 예측하였다. 결과적으로 LSTM 취수량 예측 모형은 높은 정확도를 보였으며, 월단위 전망 시 관측값에 대하여 신뢰성이 있는 결과를 나타내었다. 본 연구에서 개발한 모형에 따른 결과는 수자원 관리를 위해 활용이 가능할 것으로 기대된다.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제15권6호
• /
• pp.2168-2187
• /
• 2021

The smart grid replaces the traditional power structure with information inventiveness that contributes to a new physical structure. In such a field, malicious information injection can potentially lead to extreme results. Incorrect, FDI attacks will never be identified by typical residual techniques for false data identification. Most of the work on the detection of FDI attacks is based on the linearized power system model DC and does not detect attacks from the AC model. Also, the overwhelming majority of current FDIA recognition approaches focus on FDIA, whilst significant injection location data cannot be achieved. Building on the continuous developments in deep learning, we propose a Deep Learning based Locational Detection technique to continuously recognize the specific areas of FDIA. In the development area solver gap happiness is a False Data Detector (FDD) that incorporates a Convolutional Neural Network (CNN). The FDD is established enough to catch the fake information. As a multi-label classifier, the following CNN is utilized to evaluate the irregularity and cooccurrence dependency of power flow calculations due to the possible attacks. There are no earlier statistical assumptions in the architecture proposed, as they are "model-free." It is also "cost-accommodating" since it does not alter the current FDD framework and it is only several microseconds on a household computer during the identification procedure. We have shown that ANN-MLP, SVM-RBF, and CNN can conduct locational detection under different noise and attack circumstances through broad experience in IEEE 14, 30, 57, and 118 bus systems. Moreover, the multi-name classification method used successfully improves the precision of the present identification.

• 대한의용생체공학회:의공학회지
• /
• 제45권2호
• /
• pp.90-94
• /
• 2024

Diagnosing wounds presents a significant challenge in clinical settings due to its complexity and the subjective assessments by clinicians. Wound deep learning algorithms quantitatively assess wounds, overcoming these challenges. However, a limitation in existing research is reliance on specific datasets. To address this limitation, we created a comprehensive dataset by combining open dataset with self-produced dataset to enhance clinical applicability. In the annotation process, machine learning based on Gradient Vector Flow (GVF) was utilized to improve objectivity and efficiency over time. Furthermore, the deep learning model was equipped U-net with residual blocks. Significant improvements were observed using the input dataset with images cropped to contain only the wound region of interest (ROI), as opposed to original sized dataset. As a result, the Dice score remarkably increased from 0.80 using the original dataset to 0.89 using the wound ROI crop dataset. This study highlights the need for diverse research using comprehensive datasets. In future study, we aim to further enhance and diversify our dataset to encompass different environments and ethnicities.

• 인터넷정보학회논문지
• /
• 제21권5호
• /
• pp.21-29
• /
• 2020

Accurate estimation of human pose relies on backbone method in which its role is to extract feature map. Up to dated, the method of backbone feature extraction is conducted by the plain convolutional neural networks named by CNN and the residual neural networks named by Resnet, both of which have various architectures and performances. The CNN family network such as VGG which is well-known as a multiple stacked hidden layers architecture of deep learning methods, is base and simple while Resnet which is a bottleneck layers architecture yields fewer parameters and outperform. They have achieved inspired results as a backbone network in human pose estimation. However, they were used then followed by different pose estimation networks named by pose parsing module. Therefore, in this paper, we present a comparison between the plain CNN family network (VGG) and bottleneck network (Resnet) as a backbone method in the same pose parsing module. We investigate their performances such as number of parameters, loss score, precision and recall. We experiment them in the bottom-up method of human pose estimation system by adapted the pose parsing module of openpose. Our experimental results show that the backbone method using VGG network outperforms the Resent network with fewer parameter, lower loss score and higher accuracy of precision and recall.

• 천문학회보
• /
• 제44권1호
• /
• pp.52.1-52.1
• /
• 2019

The Helioseismic and Magnetic Imager (HMI) is the instrument of Solar Dynamics Observatory (SDO) to study the magnetic field and oscillation at the solar surface. The HMI image is not enough to analyze very small magnetic features on solar surface since it has a spatial resolution of one arcsec. Super resolution is a technique that enhances the resolution of a low resolution image. In this study, we use a method for enhancing the solar image resolution using a Deep-learning model which generates a high resolution HMI image from a low resolution HMI image (4 by 4 binning). Deep learning networks try to find the hidden equation between low resolution image and high resolution image from given input and the corresponding output image. In this study, we trained a model based on a very deep residual channel attention networks (RCAN) with HMI images in 2014 and test it with HMI images in 2015. We find that the model achieves high quality results in view of both visual and measures: 31.40 peak signal-to-noise ratio(PSNR), Correlation Coefficient (0.96), Root mean square error (RMSE) is 0.004. This result is much better than the conventional bi-cubic interpolation. We will apply this model to full-resolution SDO/HMI and GST magnetograms.