• Proceedings of the Korean Information Science Society Conference
• /
• 2005.11b
• /
• pp.868-870
• /
• 2005

스크래치는 오래된 필름에서 가장 많이 나타나는 손상 요인이다. 고화질의 멀티미디어 서비스를 제공하기 위해서는 이러한 스크래치들은 반드시 검출 및 복원되어야 한다. 이러한 중요성 때문에 지금까지 많은 복원 알고리즘이 개발되어 왔으나, 스크래치 영역의 자동검출에 대한 연구는 거의 이루어지지 않은 실정이다. 따라서 본 논문에서는 자동으로 스크래치영역을 추출할 수 있는 신경망 기반의 검출 방법을 제안한다. 다층 퍼셉트론 (Multi-layer perceptron: MLP)을 이용하여 스크래치영역과 비 스크래치영역을 구분하는데, 이 MLP는 다양한 크기의 스크래치를 검출하기 위해 다양한 크기의 입력 영상에 대해 적용된다. 제안된 방법의 평가를 위해 principal/ secondary 스크래치, alone/not-alone 스크래치, moving/static 스크래치등의 다양한 종류의 스크래치를 가진 영상에 대해 실험이 이루어졌고, 그 결과 제안된 방법의 강건함과 효율성을 입증되었다.

• The KIPS Transactions:PartB
• /
• v.16B no.2
• /
• pp.117-122
• /
• 2009

We developed a multi-lingual namecard recognizer for building up a global client management systems. At first, we created the Unicode-based character image database for character recognition and learning of multi languages, and applied many color image processing techniques to get more correct data for namecard images which were acquired by various input devices. And by applying multi-layer perceptron neural network, individual character recognition applied for language types, and post-processing utilizing keyword databases made for individual languages, we increased a recognition rate for multi-lingual namecards.

• Proceedings of the Korean Information Science Society Conference
• /
• 2000.10b
• /
• pp.305-307
• /
• 2000

본 논문에서는 다층 퍼셉트론 신경망에서 연결선 수를 최소로 하면서 일반화 성능을 향상시키기 위해 가장 널리 쓰여지고 있는 Optimal Brain Surgeon을 이용한 프루닝(pruning)을 기반으로 하여 오차 함수의 가중치 감소항을 추가시키는 방법을 사용한다. 이때 학습 및 프루닝의 성능에 많은 영향을 미치는 가중치 감소항의 방영정도를 베이시안 테크닉에 기반하여 적응적으로 최적화 하는 방법을 제안한다. 제안하는 방법의 성능을 검증하기 위해 벤치마크 데이터를 이용하여 실험을 수행하였다. 순수한 OBS 방법과 고정된 반영정도를 가진 가중치 감소항을 추가시킨 OBS, 그리고 제안하는 적응적 가중치 감소항을 적용한 OBS 방법을 비교하여 제한하는 방법이 기존의 두 방법에 비해 신경망 구조의 최적화 능력이 뛰어남을 확인할 수 있었다.

• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.14 no.3
• /
• pp.1372-1377
• /
• 2013

In this Paper, a back propagation neural network learning algorithm based on the complex multilayer perceptron is represented for controling and detecting interference of the received signals in cellular mobile communication system. We proposed neural network adaptive correlator which has fast convergence rate and good performance with combining back propagation neural network and the receiver of cellular. We analyzed and proved that NNAC has lower bit error probability than that of traditional RAKE receiver through results of computer simulation in the presence of the tone and narrow-band interference and the co-channel interference.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.8 no.6
• /
• pp.122-128
• /
• 1998

This paper presents compression of image sequences based on the classification of interframe difference image blocks. classification process consists of image activity classification and energy distribution classification. In the activity classification, interframe difference image blocks are classified into activity blocks and non-activity blocks using the edge detection. In the distribution classification, activity blocks are further classified into vertical blocks, horizontal blocks, and small activity blocks using the AC energy distribution features. The RBFN, trained with numerical classification results, successfully classifies difference image blocks according to image details. Image sequence compressing based on the classification of interframe difference image blocks using the RBFN shows better compression results and less training time than the classical sorting method and the MLP network.

• Journal of Korea Multimedia Society
• /
• v.6 no.7
• /
• pp.1312-1320
• /
• 2003

In this paper, we consider a multilayer neural network, with a single hidden layer. Error backpropagation learning method used widely in multilayer neural networks has a possibility of local minima due to the inadequate weights and the insufficient number of hidden nodes. So we propose a fuzzy supervised learning algorithm by using self-generation that self-generates hidden nodes by the compound fuzzy single layer perceptron and modified ART1. From the input layer to hidden layer, a modified ART1 is used to produce nodes. And winner take-all method is adopted to the connection weight adaptation, so that a stored pattern for some pattern gets updated. The proposed method has applied to the student identification card images. In simulation results, the proposed method reduces a possibility of local minima and improves learning speed and paralysis than the conventional error backpropagation learning algorithm.

• Proceedings of the Korea Water Resources Association Conference
• /
• 2023.05a
• /
• pp.206-206
• /
• 2023

지구온난화로 인한 기후변화에 따라 평균강수량과 증발량이 증가하며 강우지역 집중화와 강우강도가 높아질 가능성이 크다. 우리나라의 경우 협소한 국토면적과 높은 인구밀도로 기후변동의 영향이 크기 때문에 한반도에 적합한 유역규모의 수자원 예측과 대응방안을 마련해야 한다. 이를 위한 수자원 관리를 위해서는 유역에서 강수량, 유출량, 증발량 등의 장기적인 자료가 필요하며 경험식, 물리적 강우-유출 모형 등이 사용되었고, 최근들어 연구의 확장성과 비 선형성 등을 고려하기 위해 딥러닝등 인공지능 기술들이 접목되고 있다. 본 연구에서는 ASOS(동해, 태백)와 AWS(삼척, 신기, 도계) 5곳의 관측소에서 2011년~2020년까지의 일 단위 기상관측자료를 수집하고 WAMIS에서 같은 기간의 오십천 하구 일 유출량 자료를 수집 후 5개 관측소를 기준으로Thiessen 면적비를 적용해 기상자료를 구축했으며 Angstrom & Hargreaves 공식으로 잠재증발산량 산정해 3개의 모델에 각각 기상자료(일 강수량, 최고기온, 최대 순간 풍속, 최저기온, 평균풍속, 평균기온), 일 강수량과 잠재증발산량, 일 강수량 - 잠재증발산량을 학습 후 관측 유출량과 비교결과 기상자료(일 강수량, 최고기온, 최대 순간 풍속, 최저기온, 평균풍속, 평균기온)로 학습한 모델성능이 가장 높아 최적 모델로 선정했으며 일, 월, 연 관측유출량 시계열과 비교했다. 또한 같은 학습자료를 사용해 다층 퍼셉트론(Multi Layer Perceptron, MLP) 앙상블 모델을 구축하여 수자원 분야에서의 인공지능 활용성을 평가했다.

• Journal of the Institute of Electronics Engineers of Korea CI
• /
• v.41 no.3
• /
• pp.67-77
• /
• 2004

The recent research has been focused on fusion of audio and visual features for reliable speech recognition in noisy environments. In this paper, we propose a neural network based model of robust speech recognition by integrating audio, visual, and contextual information. Bimodal Neural Network(BMNN) is a multi-layer perception of 4 layers, each of which performs a certain level of abstraction of input features. In BMNN the third layer combines audio md visual features of speech to compensate loss of audio information caused by noise. In order to improve the accuracy of speech recognition in noisy environments, we also propose a post-processing based on contextual information which are sequential patterns of words spoken by a user. Our experimental results show that our model outperforms any single mode models. Particularly, when we use the contextual information, we can obtain over 90% recognition accuracy even in noisy environments, which is a significant improvement compared with the state of art in speech recognition. Our research demonstrates that diverse sources of information need to be integrated to improve the accuracy of speech recognition particularly in noisy environments.

• Industry Promotion Research
• /
• v.6 no.1
• /
• pp.1-7
• /
• 2021

We investigated multi-layer perceptrons and supervised learning algorithms, and also examined how to model functional relationships between covariates and response variables using a package called neuralnet. The algorithm applied in this paper is characterized by continuous adjustment of the weights, which are parameters to minimize the error function based on the comparison between the actual and predicted values of the response variable. In the neuralnet package, the activation and error functions can be appropriately selected according to the given situation, and the remaining parameters can be set as default values. As a result of using the neuralnet package for the infertility data, we found that age has little influence on infertility among the four independent variables. In addition, the weight of the neural network takes various values from -751.6 to 7.25, and the intercepts of the first hidden layer are -92.6 and 7.25, and the weights for the covariates age, parity, induced, and spontaneous to the first hidden neuron are identified as 3.17, -5.20, -36.82, and -751.6.

• The Journal of Engineering Geology
• /
• v.28 no.4
• /
• pp.583-592
• /
• 2018

Recently, concern has arisen regarding the lowering of groundwater levels in the hinterland caused by the development of high-capacity radial collector wells in riverbank filtration areas. In this study, groundwater levels are estimated using Modflow software in relation to the water volume pumped by the radial collector well in Anseongcheon Stream. Using the water volume data, an artificial neural network (ANN) model is developed to determine the amount of water that can be withdrawn while minimizing the reduction of groundwater level. We estimate that increasing the pumping rate of the horizontal well HW-6, which is drilled parallel to the stream direction, is necessary to minimize the reduction of groundwater levels in wells OW-7 and OB-11. We also note that the number of input data and the classification of training and test data affect the results of the ANN model. This type of approach, which supplements ANN modeling with observed data, should contribute to the future groundwater management of hinterland areas.