• Annual Conference of KIPS
• /
• 2007.11a
• /
• pp.1021-1024
• /
• 2007

인터넷이 발달하고 VoIP 의 활성화로 인하여 사용자들은 인터넷 전화서비스의 통화품질에 대한 관심이 크게 증가하고 있다. 본 고에서는 인터넷 전화서비스의 단말에 사용하고 있는 음성 코덱의 품질 측정을 위해 IP 패킷 전송품질 파라미터(ITU-T Y.1540)들을 인가하면서 이들 파라미터들의 변화에 따른 E-Model(ITU-T G.107) 종합 음성품질(R 값)을 측정하여 인터넷전화 단말의 프로세싱 품질을 파악하고 종단간서비스에서 단말이 차지하는 부분을 분석하여 인터넷 전화서비스의 통화품질 개선 및 향후 단말의 인증기준 등에 활용하고자 한다.

• Annual Conference of KIPS
• /
• 2024.05a
• /
• pp.798-800
• /
• 2024

현대의 머신러닝 환경에서는 특히 모바일 컴퓨팅 및 사물 인터넷(IoT)의 애플리케이션 영역에서 개인 정보를 보호하고 효율적이며 확장 가능한 모델에 대한 관심이 높아지고 있다. 본 연구는 연합 학습(FL)과 자기지도 학습(self-supervised learning)을 결합하여 이질적(heterogeneous)인 분산 자원에서 레이블이 없는 데이터를 활용하면서 사용자의 개인 정보를 보호하는 새로운 프레임워크를 소개한다. 이 프레임워크의 핵심은 SimCLR 과 같은 자기지도 학습 기법으로 학습된 공유 인코더로, 입력 데이터에서 고수준 특성을 추출하도록 설계되었다. 또한 이 구조를 통해 주석(annotation)이 없는 방대한 데이터셋을 활용하여 모델 성능을 향상시키고, 여러 개의 격리된 모델이 필요하지 않아 리소스를 크게 최적화할 수 있는 가능성을 확인했다. 본 연구를 통해 생성된 모델은 중앙 집중 방식(CL)이면서 자기지도학습으로 학습되지 않은 기존 모델과 비교하여 전체 평균 정확도가 14.488% 향상됐다.

• Annual Conference of KIPS
• /
• 2024.05a
• /
• pp.321-322
• /
• 2024

다양한 방면에서 사용되는 임베디드 시스템의 메모리 취약성에 대한 관심이 많아짐에 따라 임베디드 시스템의 메모리 보호와 관련하여 많은 연구가 진행 중이다. 스택 카나리는 효율적인 메모리 보호 기법으로써 널리 사용되지만 물리 메모리가 제한적이고 사용자 권한 분리를 지원하지 않는 임베디드 시스템에서는 기존 방식을 활용한 스택 카나리를 적용하는 것에 한계가 있다. ARM의 TrustZone은 일반 실행 환경과 신뢰 실행 환경으로 분리하여 일반 실행 환경에서 신뢰 실행 환경의 데이터나 코드에 접근하지 못 하도록 막는다. 그렇기 때문에 ARM의 TrustZone에 암호화 키를 저장하거나 보안이 중요한 동작을 TrustZone에서 실행하는 연구가 많다. 본 논문에서는 ARM의 TrustZone을 활용하여 임베디드 시스템에서 스택 카나리 기법의 한계를 보완 할 수 있는 방법을 제안한다.

• Annual Conference of KIPS
• /
• 2024.10a
• /
• pp.932-933
• /
• 2024

AI 로보어드바이저를 활용한 금융 추천 서비스 'PURE'는 사용자의 투자 성향과 관심 분야를 분석하여 맞춤형 종목을 추천하고, ChatGPT API로 트렌드 및 감성 분석을 통해 기사 요약 기능을 제공한다. 최신 AI 기술과 빅데이터를 활용해 개인화된 투자 조언과 리스크 관리를 지원하며, 실시간 데이터 분석을 통해 투자 전략을 강화하고 비용 절감 및 시장 변화에 유연하게 대응한다.

• Annual Conference of KIPS
• /
• 2024.10a
• /
• pp.1049-1050
• /
• 2024

최근 패션 소비자들의 지속 가능한 소비와 환경 보호에 대한 관심이 높아지면서, 기존 AI 의류 추천 시스템의 과소비 촉진 문제가 부각되고 있다. 본 논문에서는 사용자가 소유한 의류를 기반으로 스타일을 제안하는 패션 큐레이션 시스템을 제안하며, AWS Rekognition을 통해 의류의 시각적 특징을 분석하고 날씨와 TPO 정보를 반영한 맞춤형 추천을 제공한다. 또한, 타 서비스와 차별화된 '시밀러룩' 추천 기능을 도입하여, 주변인과 자연스럽게 어울리는 스타일을 추구할 수 있도록 돕는다. 이는 환경 보호와 윤리적 소비를 촉진하며, 패션 산업의 과소비 문제 해결에 기여할 것으로 기대된다.

• Annual Conference of KIPS
• /
• 2024.10a
• /
• pp.424-426
• /
• 2024

인공지능 기술이 급속도로 발전하고 있음에도 암환자의 방사선 치료 계획시 종양 경계를 임상의가 여전히 수동으로 설정하고 있다. 본 연구에서는 2 차원 및 3 차원 의료영상에서 종양 및 정상 조직을 자동으로 분할하고 시각화 해주는 소프트웨어 도구를 소개하고자 한다. 개발된 소프트웨어는 사용자의 개입을 최소화하고, 분할하고자 하는 관심 영역에 따라 파라미터 설정을 달리할 수 있다. 오픈 소프트웨어 3DSlicer 와 종양의 활성, 괴사, 부종 영역에 대한 분할 정확성을 DSC(Dice similarity coefficient)로 평가했으며, 비슷하거나 더 우수한 결과를 보였다. 특히, 초기 컨투어 설정부터 최종 분할 마스크 획득까지 처리 시간이 약 36.3% 더 빠른 결과를 보였다. 개발된 의료영상 분할 소프트웨어는 종양 및 정상 조직 분할을 용이하게 하여, 인공지능 지도 학습에 필요로 하는 라벨링 작업을 지원할 수 있을 것으로 사료된다.

• 한국HCI학회:학술대회논문집
• /
• 2009.02a
• /
• pp.826-833
• /
• 2009

Currently, the rise of social tagging has changing taxonomy to folksonomy. Tag represents a new approach to organizing information. Nonhierarchical classification allows data to be freely gathered, allows easy access, and has the ability to move directly to other content topics. Tag is expected to play a key role in clustering various types of contents, it is expand to network in the common interests among users. First, this paper determine the relationships among user, tags and resources in social tagging system and examine the circumstances of what aspects to users when creating a tag related to features of websites. Therefore, this study uses tags from the social bookmarking service 'del.icio.us' to analyze the features of tag words when adding a new web page to a list. To do this, websites features classified into 7 items, it is known as tag classification related to resources. Experiments were conducted to test the proposed classify method in the area of music, photography and games. This paper attempts to investigate the perspective in which users apply a tag to a webpage and establish the capacity of expanding a social service that offers the opportunity to create a new business model.

• Journal of the Institute of Electronics and Information Engineers
• /
• v.52 no.11
• /
• pp.105-114
• /
• 2015

This paper proposes methods to remove background area and segment region of interest (ROI) in chest X-ray images. Conventional algorithms to improve detail or contrast of images normally utilize brightness and frequency information. If we apply such algorithms to the entire images, we cannot obtain reliable visual quality due to unnecessary information such as background area. So, we propose two effective algorithms to remove background and segment ROI from the input X-ray images. First, the background removal algorithm analyzes the histogram distribution of the input X-ray image. Next, the initial background is estimated by a proper thresholding on histogram domain, and it is removed. Finally, the body contour or background area is refined by using a popular guided filter. On the other hand, the ROI, i.e., lung segmentation algorithm first determines an initial bounding box using the lung's inherent location information. Next, the main intensity value of the lung is computed by vertical cumulative sum within the initial bounding box. Then, probable outliers are removed by using a specific labeling and the pre-determined background information. Finally, a bounding box including lung is obtained. Simulation results show that the proposed background removal and ROI segmentation algorithms outperform the previous works.

• Journal of Intelligence and Information Systems
• /
• v.22 no.2
• /
• pp.81-96
• /
• 2016

As the internet has become widespread and easy to access everywhere, it is common for people to search information via online search engines such as Google and Naver in everyday life. Recent studies have used online search volume of specific keyword as a measure of the internet users' attention in order to predict disease outbreaks such as flu and cancer, an unemployment rate, and an index of a nation's economic condition, and etc. For stock traders, web search is also one of major information resources to obtain data about individual stock items. Therefore, search volume of a stock item can reflect the amount of investors' attention on it. The investor attention has been regarded as a crucial factor influencing on stock price but it has been measured by indirect proxies such as market capitalization, trading volume, advertising expense, and etc. It has been theoretically and empirically proved that an increase of investors' attention on a stock item brings temporary increase of the stock price and the price recovers in the long run. Recent development of internet environment enables to measure the investor attention directly by the internet search volume of individual stock item, which has been used to show the attention-induced price pressure. Previous studies focus mainly on Dow Jones and NASDAQ market in the United States. In this paper, we investigate the relationship between the individual investors' attention measured by the internet search volumes and stock price changes of individual stock items in the KOSDAQ market in Korea, where the proportion of the trades by individual investors are about 90% of the total. In addition, we examine the difference between industries in the influence of investors' attention on stock return. The internet search volume of stocks were gathered from "Naver Trend" service weekly between January 2007 and June 2015. The regression model with the error term with AR(1) covariance structure is used to analyze the data since the weekly prices in a stock item are systematically correlated. The market capitalization, trading volume, the increment of trading volume, and the month in which each trade occurs are included in the model as control variables. The fitted model shows that an abnormal increase of search volume of a stock item has a positive influence on the stock return and the amount of the influence varies among the industry. The stock items in IT software, construction, and distribution industries have shown to be more influenced by the abnormally large internet search volume than the average across the industries. On the other hand, the stock items in IT hardware, manufacturing, entertainment, finance, and communication industries are less influenced by the abnormal search volume than the average. In order to verify price pressure caused by investors' attention in KOSDAQ, the stock return of the current week is modelled using the abnormal search volume observed one to four weeks ahead. On average, the abnormally large increment of the search volume increased the stock return of the current week and one week later, and it decreased the stock return in two and three weeks later. There is no significant relationship with the stock return after 4 weeks. This relationship differs among the industries. An abnormal search volume brings particularly severe price reversal on the stocks in the IT software industry, which are often to be targets of irrational investments by individual investors. An abnormal search volume caused less severe price reversal on the stocks in the manufacturing and IT hardware industries than on average across the industries. The price reversal was not observed in the communication, finance, entertainment, and transportation industries, which are known to be influenced largely by macro-economic factors such as oil price and currency exchange rate. The result of this study can be utilized to construct an intelligent trading system based on the big data gathered from web search engines, social network services, and internet communities. Particularly, the difference of price reversal effect between industries may provide useful information to make a portfolio and build an investment strategy.

• Journal of Intelligence and Information Systems
• /
• v.18 no.3
• /
• pp.97-117
• /
• 2012

Nowadays, Internet users tend to do a variety of actions at the same time such as web browsing, social networking and multimedia consumption. While watching a video, once a user is interested in any product, the user has to do information searches to get to know more about the product. With a conventional approach, user has to search it separately with search engines like Bing or Google, which might be inconvenient and time-consuming. For this reason, a video annotation platform has been developed in order to provide users more convenient and more interactive ways with video content. In the future of smart TV environment, users can follow annotated information, for example, a link to a vendor to buy the product of interest. It is even better to enable users to search for information by directly discussing with friends. Users can effectively get useful and relevant information about the product from friends who share common interests or might have experienced it before, which is more reliable than the results from search engines. Social networking services provide an appropriate environment for people to share products so that they can show new things to their friends and to share their personal experiences on any specific product. Meanwhile, they can also absorb the most relevant information about the product that they are interested in by either comments or discussion amongst friends. However, within a very huge graph of friends, determining the most appropriate persons to ask for information about a specific product has still a limitation within the existing conventional approach. Once users want to share or discuss a product, they simply share it to all friends as new feeds. This means a newly posted article is blindly spread to all friends without considering their background interests or knowledge. In this way, the number of responses back will be huge. Users cannot easily absorb the relevant and useful responses from friends, since they are from various fields of interest and knowledge. In order to overcome this limitation, we propose a method to filter a user's friends for information search, which leverages semantic video annotation and social networking services. Our method filters and brings out who can give user useful information about a specific product. By examining the existing Facebook information regarding users and their social graph, we construct a user profile of product interest. With user's permission and authentication, user's particular activities are enriched with the domain-specific ontology such as GoodRelations and BestBuy Data sources. Besides, we assume that the object in the video is already annotated using Linked Data. Thus, the detail information of the product that user would like to ask for more information is retrieved via product URI. Our system calculates the similarities among them in order to identify the most suitable friends for seeking information about the mentioned product. The system filters a user's friends according to their score which tells the order of whom can highly likely give the user useful information about a specific product of interest. We have conducted an experiment with a group of respondents in order to verify and evaluate our system. First, the user profile accuracy evaluation is conducted to demonstrate how much our system constructed user profile of product interest represents user's interest correctly. Then, the evaluation on filtering method is made by inspecting the ranked results with human judgment. The results show that our method works effectively and efficiently in filtering. Our system fulfills user needs by supporting user to select appropriate friends for seeking useful information about a specific product that user is curious about. As a result, it helps to influence and convince user in purchase decisions.