• Spatial Information Research
• /
• 제17권4호
• /
• pp.405-416
• /
• 2009

도시의 패러다임이 변화하고 이에 따라 u-City의 수요가 급증하고 있다. 우리나라에서만 2009년 5월을 기준으로 54개 지역에서 u-City가 구축되고 있다. 이렇게 급증하는 u-City의 성패를 좌우하는 중요 요소 중 하나가 바로 u-City에서 제공하고자 하는 u-서비스라고 할 수 있겠다. 현재의 u-서비스는 센서네트워크를 기반으로 하는 도시 관리 위주의 모니터링 서비스가 대부분이다. 유비쿼터스의 본질적 의미인 '언제 어디서나'의 u-서비스를 구현하기 위해서는 공간정보의 활용이 필수적이다. 공간정보와 센서정보의 결합은 공간인지(spatial awareness) 를 가능케 하여, 모니터링서비스에서 보다 나아가 공간분석이 가능한 서비스로의 발전을 도모할 수 있다. 따라서 본 연구에서는 1) u-서비스에서 spatial awareness는 어떤 의미 인지를 명확히 하고, 2) spatial awareness를 가능케 하기 위해 공간정보, 센서정보, 기타 정보들이 어떻게 융합 (Spatial Embedding) 되어야 하는지 논하고, 3) U-GIS를 기반으로 한 u-서비스 시나리오 4가지를 제안하고, 4) 현재의 기술개발 현황에 대해 분석해 보고자 한다.

• 한국정보통신학회논문지
• /
• 제18권12호
• /
• pp.2946-2952
• /
• 2014

표본 기반 초해상도(Super Resolution 이하 SR) 방법들 중 네이버 임베딩(Neighbor Embedding 이하 NE) 기법의 기본 원리는 지역적 선형 임베딩이라는 매니폴드 학습방법의 개념과 같다. 그러나, 네이버 임베딩은 국부 학습 데이터 집합의 크기가 너무 작기 때문에 이에 따른 빈약한 일반화 능력으로 인하여 알고리즘의 성능을 크게 저하시킨다. 본 논문에서는 이와 같은 문제점을 해결하기 위해서 일반화 능력이 뛰어난 Support Vector Regression(이하 SVR)을 이용한 Sparse-Neighbor 영상 표현 학습 방법에 기반한 새로운 알고리즘을 제안하였다. 저해상도 입력 영상이 주어지면 bicubic 보간법을 이용하여 확대된 영상을 얻고, 이 확대된 영상으로부터 패치를 얻은 후 저주파 패치인지 고주파 패치 인지를 판별한 후 각 영상 패치의 가중치를 얻은 후 두 개의 SVR을 훈련하였으며 훈련된 SVR을 이용하여 고해상도의 해당 화소 값을 예측하였다. 실험을 통하여 제안된 기법이 기존의 보간법 및 네이버 임베딩 기법 등에 비해 정량적인 척도 및 시각적으로 향상된 결과를 보여 주었다.

• 한국ITS학회 논문지
• /
• 제20권1호
• /
• pp.160-173
• /
• 2021

최근 전 세계적으로 자율주행자동차 개발을 위한 연구가 증가하고 있으며, 자율주행자동차의 실도로 도입이 증가되고 있는 추세이다. 하지만, 자율주행자동차의 교통사고 발생으로 인해 자율주행자동차 안전성에 대한 관심이 높아지고 있다. 또한, 자율주행자동차 교통사고에 대한 특성 파악 및 분석 방법론 개발의 필요성이 대두되고 있다. 특히 미국 캘리포니아 차량관리국(California Department of Motor Vehicles, DMV)에서는 자율주행자동차의 교통사고 데이터를 수집하여 리포트 형태로 제공하고 있다. 본 연구에서는 DMV에서 제공하는 자율주행자동차 교통사고를 분석하는 방법론을 제시하였다. 또한, 텍스트 임베딩 기법을 이용하여 주요 키워드 및 주요 토픽 도출을 통해 개발된 방법론의 활용도를 검토하였다. 본 연구에서 개발된 방법론은 향후 자율주행자동차 교통사고 데이터가 충분히 수집된다면 자율주행자동차 교통사고 분석 및 자율주행자동차 개발시 활용될 수 있을 것으로 기대된다.

• Journal of Information Processing Systems
• /
• 제13권3호
• /
• pp.599-617
• /
• 2017

Due to the rapid growth and expansion of the Internet, the digital multimedia such as image, audio and video are available for everyone. Anyone can make unauthorized copying for any digital product. Accordingly, the owner of these products cannot protect his ownership. Unfortunately, this situation will restrict any improvement which can be done on the digital media production in the future. Some procedures have been proposed to protect these products such as cryptography and watermarking techniques. Watermarking means embedding a message such as text, the image is called watermark, yet, in a host such as a text, an image, an audio, or a video, it is called a cover. Watermarking can provide and ensure security, data authentication and copyright protection for the digital media. In this paper, a new watermarking method of still image is proposed for the purpose of copyright protection. The procedure of embedding watermark is done in a transform domain. The discrete cosine transform (DCT) is exploited in the proposed method, where the watermark is embedded in the selected coefficients according to several criteria. With this procedure, the deterioration on the image is minimized to achieve high invisibility. Unlike the traditional techniques, in this paper, a new method is suggested for selecting the best blocks of DCT coefficients. After selecting the best DCT coefficients blocks, the best coefficients in the selected blocks are selected as a host in which the watermark bit is embedded. The coefficients selection is done depending on a weighting function method, where this function exploits the values and locations of the selected coefficients for choosing them. The experimental results proved that the proposed method has produced good imperceptibility and robustness for different types of attacks.

• 지능정보연구
• /
• 제5권1호
• /
• pp.103-123
• /
• 1999

Input filtering as a preprocessing method is so much crucial to get good performance in time series forecasting. There are a few preprocessing methods (i.e. ARMA outputs as time domain filters, and Fourier transform or wavelet transform as time-frequency domain filters) for handling time series. Specially, the time-frequency domain filters describe the fractal structure of financial markets better than the time domain filters due to theoretically additional frequency information. Therefore, we, first of all, try to describe and analyze specially some issues on the effectiveness of different filtering methods from viewpoint of the performance of a neural network based forecasting. And then we discuss about neural network model architecture issues, for example, what type of neural network learning architecture is selected for our time series forecasting, and what input size should be applied to a model. In this study an input selection problem is limited to a size selection of the lagged input variables. To solve this problem, we simulate on analyzing and comparing a few neural networks having different model architecture and also use an embedding dimension measure as chaotic time series analysis or nonlinear dynamic analysis to reduce the dimensionality (i.e. the size of time delayed input variables) of the models. Throughout our study, experiments for integration methods of joint time-frequency analysis and neural network techniques are applied to a case study of daily Korean won / U. S dollar exchange returns and finally we suggest an integration framework for future research from our experimental results.

• 한국컴퓨터정보학회논문지
• /
• 제29권8호
• /
• pp.53-58
• /
• 2024

교통사고는 인간의 생명뿐만 아니라 사회적으로 큰 비용을 발생시키는 문제이다. 최근에는 교통사고 문제를 해결하기 위하여, 딥러닝 기술과 도로의 시공간적 정보를 통해 교통사고 위험도를 예측하는 연구가 진행되었다. 그러나 교통사고는 도로의 시공간적 정보뿐만 아니라 인적요소 또한 교통사고에 매우 큰 영향을 미치지만 이에 대한 연구는 상대적으로 활성화되지 않았다. 본 논문은 교통사고 데이터셋을 바탕으로 클러스터링 기법을 적용하여 운전자 그룹 및 특성을 분석하였으며, 각 운전자 그룹 및 특성에 대한 위험도를 산출하는 방법을 제시 및 적용하였다. 이 과정에서 본 논문에서 제시한 전처리 기법이 기존에 일반적으로 사용되었던 원-핫 임베딩, Min-Max Scaling 기법보다 더 높은 성능을 보임으로써 더 적합한 전처리 기법임을 보였다.

• 한국복합재료학회:학술대회논문집
• /
• 한국복합재료학회 2005년도 춘계학술발표대회 논문집
• /
• pp.17-20
• /
• 2005

In-situ structural health monitoring of filament wound pressure tanks were conducted during water-pressurizing test using embedded fiber Bragg grating (FBG) sensors. We need to monitor inner strains during working in order to verify the health condition of pressure tanks more accurately because finite element analyses on filament wound pressure tanks usually show large differences between inner and outer strains. Fiber optic sensors, especially FBG sensors can be easily embedded into the composite structures contrary to conventional electric strain gages (ESGs). In addition, many FBG sensors can be multiplexed in single optical fiber using wavelength division multiplexing (WDM) techniques. We fabricated a standard testing and evaluation bottle (STEB) with embedded FBG sensors and performed a water-pressurizing test. In order to increase the survivability of embedded FBG sensors, we suggested a revised fabrication process for embedding FBG sensors into a filament wound pressure tank, which includes a new protecting technique of sensor heads, the grating parts. From the experimental results, it was demonstrated that FBG sensors can be successfully adapted to filament wound pressure tanks for their structural health monitoring by embedding.

• 한국정보과학회 언어공학연구회:학술대회논문집(한글 및 한국어 정보처리)
• /
• 한국정보과학회언어공학연구회 2018년도 제30회 한글 및 한국어 정보처리 학술대회
• /
• pp.652-657
• /
• 2018

With the rapid development of e-commerce, many customers can now express their opinion on various kinds of product at discussion groups, merchant sites, social networks, etc. Discerning a consensus opinion about a product sold online is difficult due to more and more reviews become available on the internet. Opinion Mining, also known as Sentiment analysis, is the task of automatically detecting and understanding the sentimental expressions about a product from customer textual reviews. Recently, researchers have proposed various approaches for evaluation in sentiment mining by applying several techniques for document, sentence and aspect level. Aspect-based sentiment analysis is getting widely interesting of researchers; however, more complex algorithms are needed to address this issue precisely with larger corpora. This paper introduces an approach of knowledge representation for the task of analyzing product aspect rating. We focus on how to form the nature of sentiment representation from textual opinion by utilizing the representation learning methods which include word embedding and compositional vector models. Our experiment is performed on a dataset of reviews from electronic domain and the obtained result show that the proposed system achieved outstanding methods in previous studies.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제14권10호
• /
• pp.4042-4059
• /
• 2020

The development of new multimedia techniques such as 3D printing is increasingly attracting the public's attention towards 3D objects. An optimized robust and imperceptible watermarking method based on Ant Colony Optimization (ACO) and Weber Law is proposed for 3D polygonal models. The proposed approach partitions the host model into smaller sub meshes and generates a secret watermark from the sub meshes using Weber Law. ACO based optimized strength factor is identified for embedding the watermark. The secret watermark is embedded and extracted on the wavelet domain. The proposed scheme is robust against geometric and photometric attacks that overcomes the synchronization problem and authenticates the secret watermark from the distorted models. The primary characteristic of the proposed system is the flexibility achieved in data embedding capacity due to the optimized strength factor. Extensive simulation results shows enhanced performance of the recommended framework and robustness towards the most common attacks like geometric transformations, noise, cropping, mesh smoothening, and the combination of such attacks.

• 융합신호처리학회논문지
• /
• 제15권1호
• /
• pp.1-8
• /
• 2014

Digital watermarking techniques are attracting attention as a proper solution to protect copyright for multimedia data. This paper proposes a new audio watermarking method based on Discrete Cosine Transformation (DCT) and Discrete Wavelet Transformation (DWT) for copyright protection. In our proposed watermarking method, the original audio is transformed into DCT domain and divided into two parts. Synchronization code is applied on the signal in first part and 2 levels DWT domain is applied on the signal in second part. The absolute value of DWT coefficient is divided into arbitrary number of segments and calculates the energy of each segment and middle peak. Watermarks are then embedded into each middle peak. Watermarks are extracted by performing the inverse operation of watermark embedding process. Experimental results show that the hidden watermark data is robust to re-sampling, low-pass filtering, re-quantization, MP3 compression, cropping, echo addition, delay, and pitch shifting, amplitude change. Performance analysis of the proposed scheme shows low error probability rates.