• Proceedings of the Acoustical Society of Korea Conference
• /
• autumn
• /
• pp.81-84
• /
• 2001

본 논문은 중규모 단어급의 핵심어 검출 시스템에서 인식률 향상을 위해 미등록어 거절(Out-of-Vocabulary rejection) 기능을 제어하기 위한 연구이다. 이것은 핵심어 검출기에서 인식된 결과를 확인하는 과정으로 검증시스템이 구현되기 위해서는 매 음소마다 검증 기능이 필요하고, 이를 위해서 반음소(anti-phoneme model) 모델을 사용하였다. 검증의 역할은 인식기에서 인식된 단어가 등록어인지 미등록어인지 판별하는 것이다. 단어인식기는 비터비 탐색을 하므로, 기본적으로 단어단위로 인식을 하지만 그 인식된 단어는 내부적으로 음소단위로 인식된다. 따라서, 최소 검증 오류를 갖는 반음소 모델을 사용하고, 이를 이용하여 인식된 음소 단위들을 각각의 반음소 모델과 비교하여 통계적인 방법에 의해 신뢰도를 구한다 이 음소단위의 신뢰도를 단어 단위의 신뢰도로 환산하기 위해서 음소단위를 평균 내는 방식 을 취한다. 이렇게 함으로서, 등록어와 미등록어 사이의 분별력을 크게 하여 향상된 인식 성능을 얻었다.

• Proceedings of the Korea Contents Association Conference
• /
• 2017.05a
• /
• pp.311-312
• /
• 2017

이 논문에서는 SNS 스트림을 분석하여 산림재해 (산불, 산사태 등)에 관련한 메시지를 검출하는 시스템을 구현한다. 구현하는 시스템은 SNS 스트림에서 사전에 부여된 산림 재해 과련 키워드들을 포함하는 메시지를 실시간으로 검색한다. 검색 결과에 대해 산림재해와의 관련성을 관리자가 분류를 하고 분류한 결과는 기계학습 기법을 통해 학습되어 보다 정확한 산림재해 메시지 검출을 가능하게 한다. 제안하는 방법은 Spark Streaming 과 MLIB를 기반으로 구현한다.

• Annual Conference on Human and Language Technology
• /
• 1992.10a
• /
• pp.545-553
• /
• 1992

문자인식에 있어서 인식율을 높이기 위한 후처리의 한 방법으로서, 문법 정보를 이용하는 후처리를 제안하고자 한다. 즉, 문자 인식 시스템에 의해 인식된 국어문에 대해서 오인식된 문자를 포함하는 어절을 검출하고, 오인식된 문자의 적절한 후보를 선정하여 그에 따라 자동수정을 행하는 것을 전채 후처리 과정으로 전제한다. 본 논문에서는 형태소 분석을 통해 오인식된 부분을 검출하는 과정을 보임으로써 문자인식에 있어서 문법 정보를 이용하는 후처리의 가능성과 그 유효성을 보이는 것을 목적으로 한다.

• Parasites, Hosts and Diseases
• /
• v.32 no.3
• /
• pp.163-170
• /
• 1994

An epidemiological survey was performed to know the infection status of heterophyid fluke metacercariae in the brackish water fishes purchased from a local market in Haenam-gun, Chollanam-do, Korea from November to December, 1991. A total 20 out of 30 shads (Konosiw punctatus) examined were infected with the metacercariae of H. continua (average number of larvae per an infected fish: 2.71). Of 20 perches (Lateolabroxjoponicurs), 19 were Infected with the metacercariae of H. continua (average number per an infected fish: 17.21. Of 30 mullets (Mugil cepholus) examined, 11 were Infected with the metacercariae of H. nocens (average number per an Infected fish: 51.5), 10 mullets harboured 31 and 349 metacercariae of f summa and 5. lucatum respectively. Of 30 gobies (Acnnthogobin jlcvimonus) examined, 3 were infEcted with the metacercariae of H. nucens (average number per an Infected fish: 32.0), 8 with those of H. continua (average number per an infected flesh: 12.6) and 3 with those of 5. JuscntuwL From the above results, It has been confirmed that the brackish water Ash from Haenam-gun, Chollanam-do are infected with numerous metacercarlae of H. nocens, H. continua, P. summo and S. fuscatum.

• Korean Journal of Heritage: History & Science
• /
• v.42 no.3
• /
• pp.132-153
• /
• 2009

A lotus flower pattern of the shape of straight and pointed petals has the forms whose lotus flowers are pointed, and is one of the representative forms of lotus flower pattern of Baekjae along with a round and a triangle protruded style. This style of lotus flower pattern was applied to roof-end tiles, halos of Buddhist images, containers of Buddha's bones, head supports, etc. from the Hanseong period to the Sabi period of Baekjae. However most of lotus flower patterns that remain today were used for the tiles of the Sabi period. Many of them were produced under the influence of the Northern Dynasties and the Southern Dynasties of China, and of Goryeo, which implies the active cultural exchange of Baekjae at that time. Among the present lotus flower patterns of the shape of the straight and pointed petals, that of the earliest time is from Gyeongdang district of Poongnab mud castle and belongs to the mid-fifth century. However there is a higher chance that the gradual subsequent excavation and research will find some tiles of the earlier period and other styles can also be unearthed.

• Journal of the Institute of Electronics and Information Engineers
• /
• v.51 no.8
• /
• pp.114-125
• /
• 2014

In a conventional sonar system, which uses LFM signal for detecting targets with varying speed, the results of multiple LFM Doppler correlators are aligned and the maximum alined result are selected as a test cell for detecting targets. As the number of the LFM Doppler correlators are increased for accurate target detection, as the required computational complexity and the memory are also increased. This fact makes it difficult to implement the accurate LFM target detector. In this paper, we propose a new fast target detection which is robust for the variation of target speed. Because the proposed method uses the summation of alined results of large numbers of LFM Doppler correlators, the proposed method increase SNR and provide robust SNR for the variation of target speed. And the proposed method can provide very fast target detection by implementing the process, the summation of alined results of large numbers of LFM Doppler correlators, as one summation filter.

• Journal of the Institute of Convergence Signal Processing
• /
• v.16 no.2
• /
• pp.44-49
• /
• 2015

When the scene has less texture or when camera pose largely changes, the existing texture-based feature tracking methods are not reliable. Topology-based feature description methods, which use the geometric relationship between features such as LLAH, is a good alternative. However, they require feature detection methods with high performance. As a basic study on developing an effective feature detection method for topology-based feature description, this paper aims at examining their applicability to topology-based feature description by analyzing the repeatability of several feature detection methods that are included in the OpenCV library. Experimental results show that FAST outperforms the others.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2022.05a
• /
• pp.97-99
• /
• 2022

Recently, the amount of data is increasing with the development of industries and technologies. Research on the processing and transmission of large amounts of data is attracting attention. Therefore, in this paper, compressed sensing was used to reduce the amount of data and its effect on Mask R-CNN algorithm was analyzed. We confirmed that as the compressed sensing rate increases, the amount of data in the image and the resolution decreases. However, it was confirmed that there was no significant degradation in the performance of object detection.

• Journal of Broadcast Engineering
• /
• v.18 no.3
• /
• pp.393-400
• /
• 2013

Hand shape recognition is a fundamental technique for implementing natural human-computer interaction. In this paper, we discuss a method for effectively detecting a hand region in Kinect-based hand shape recognition. Since Kinect is a camera that can capture color images and infrared images (or depth images) together, both images can be exploited for the process of detecting a hand region. That is, a hand region can be detected by finding pixels having skin colors or by finding pixels having a specific depth. Therefore, after analyzing the performance of each, we need a method of properly combining both to clearly extract the silhouette of hand region. This is because the hand shape recognition rate depends on the fineness of detected silhouette. Finally, through comparison of hand shape recognition rates resulted from different hand region detection methods in general environments, we propose a high-performance hand region detection method.

• Journal of the Korea Society of Computer and Information
• /
• v.14 no.11
• /
• pp.179-186
• /
• 2009

In this paper, we propose a head detection method based on vertical and horizontal pixel histogram analysis in order to overcome drawbacks of the previous head detection approach using Haar-like feature-based face detection. In the proposed method, we create the vertical and horizontal foreground pixel histogram images from the background subtraction image, which represent the number of foreground pixels in the same vertical or horizontal position. Then we extract feature points of a head region by applying Harris corner detection method to the foreground pixel histogram images and by analyzing corner points. The proposal method shows robust head detection results even in the face image covering forelock by hairs or the back view image in which the previous approaches cannot detect the head regions.