• Proceedings of the Korean Information Science Society Conference
• /
• 2001.04b
• /
• pp.298-300
• /
• 2001

본 논문에서는 구조적응 자기구성 지도 모델을 사용하여 인간 행동의 성별을 분류하는 인식기를 제안하였다. 26명의 사람이 '화난 상태' 혹은 '보통 상태'의 두가지 정서 하에서 '문 두드리기', '손 흔들기', '물건 들어올리기'의 세가지 동작을 수행하는 동안, 행위자 관절점의 속도나 위치 정보로부터 성별을 분류하였다. 또한 SASOM의 성능 비교 분석을 위하여 전통적인 SOM, 다층 퍼셉트론과 거의 두 가지 결합 모델, SASOM와 의사결정트리 결합 모델, 단일 의사 결정트리, $textsc{k}$-최근접 이웃 등의 인식기를 구현하여 성능을 비교분석 하였다. 실험 결과 SASOM 분류기가 가장 높은 이식률을 보였으며 분류기로서 유용함을 알 수 있었다.

• Proceedings of the Korean Information Science Society Conference
• /
• 2002.04b
• /
• pp.502-504
• /
• 2002

본 논문에서는 대량의 말뭉치에서 추출된 잠재의미에 기반하여 단어간 유사도를 측정하고 이를 영한 기계 번역에서의 역어선택에 적용한다. 잠재의미 추출을 위해서는 latent semantic analysis(LSA)와 probabilistic LSA(PLSA)를 이용한다. 주어진 단어의 역어 선택시 기본적으로 연어(collocation) 사전을 검색하고, 미등록 단어의 경우 등재된 단어 중 해당 단어와 유사도가 높은 항목의 정보를 활용하며 이 때 $textsc{k}$-최근접 이웃 방법이 이용된다. 단어들간의 유사도 계산은 잠재의미 공간상에서 이루어진다. 실험에서, 연어사전만 이용하였을 경우보다 최고 15%의 성능 향상을 보였으며, PLSA에 기반한 방법이 LSA에 의한 방법보다 역어선택 성능 면에서 약간 더 우수하였다.

• Proceedings of the Korea Inteligent Information System Society Conference
• /
• 2007.11a
• /
• pp.324-328
• /
• 2007

본 논문은 감시정찰 센서네트워크에서 센서노드의 주요 기능인 표적의 탐지 및 식별을 위한 알고리즘을 제안한다. 감시정찰 센서네트워크에서 각 센서노드는 노드의 크기 및 센서, 프로세서, 네트워크, 전원 등의 자원의 제약이 있기 때문에 침입하는 적의 탐지 및 종류 식별을 위해서는 효율적인 알고리즘의 선정과 최적화가 요구된다. 본 논문에서는 음향, 진동, PIR, 자기 센서 등을 이용하여 사람, 차량 및 궤도 차량의 침입을 탐지하기 위한 적응 임계값 알고리즘과 그 종류를 식별하기 위한 최대우도추정 기법, k-최근접 이웃 추정 기법에 기반한 표적의 탐지 및 식별 알고리즘을 제안한다. 실험결과 음향 및 진동 센서에 의한 차량의 탐지, PIR 센서에 의한 사람의 탐지가 가능함을 확인할 수 있었으며 주파수 특징점을 이용하여 차량과 궤도차량의 종류식별이 가능함을 확인할 수 있었다.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2021.11a
• /
• pp.470-473
• /
• 2021

본 논문에서는 코로나-19로 인해 홈 트레이닝 시장이 성장하고 있는 상황 속에서 효율적인 운동을 위해 사용자의 식습관, 신체조건, 선호도 등을 바탕으로 적합한 운동을 추천해주는 시스템을 제안한다. 먼저 K-최근접 이웃 알고리즘을 활용해 비만의 정도에 따라 사용자를 분류하고, 운동 데이터를 소모 칼로리에 따라 클러스터링 한다. 다음으로 비만의 정도와 운동 레벨에 따라 정해진 추천 점수를 통해 사전 선호도 확률을 계산하고, 베이지안 네트워크를 통해 사후 확률을 구한다. 이를 바탕으로 특이값 분해 알고리즘(SVD)를 활용하여 사용자 맞춤형 운동을 추천한다. 제안 시스템의 성능을 검증하기 위해 비교 실험을 진행하여 회귀 문제 평가 척도인 RMSE 값 측면에서 성능을 분석하였다.

• Journal of Broadcast Engineering
• /
• v.26 no.3
• /
• pp.283-295
• /
• 2021

This paper proposes a method for generating dense depth map using information of color images and depth map generated based on lidar based on self-organizing map. The proposed depth map upsampling method consists of an initial depth prediction step for an area that has not been acquired from LiDAR and an initial depth filtering step. In the initial depth prediction step, stereo matching is performed on two color images to predict an initial depth value. In the depth map filtering step, in order to reduce the error of the predicted initial depth value, a self-organizing map technique is performed on the predicted depth pixel by using the measured depth pixel around the predicted depth pixel. In the process of self-organization map, a weight is determined according to a difference between a distance between a predicted depth pixel and an measured depth pixel and a color value corresponding to each pixel. In this paper, we compared the proposed method with the bilateral filter and k-nearest neighbor widely used as a depth map upsampling method for performance comparison. Compared to the bilateral filter and the k-nearest neighbor, the proposed method reduced by about 6.4% and 8.6% in terms of MAE, and about 10.8% and 14.3% in terms of RMSE.

• Asia-pacific Journal of Multimedia Services Convergent with Art, Humanities, and Sociology
• /
• v.8 no.10
• /
• pp.63-70
• /
• 2018

Biomedical signal measurement technology using images has been developed, and researches on respiration signal measurement technology for maintaining life have been continuously carried out. The existing technology measured respiratory signals through a thermal imaging camera that measures heat emitted from a person's body. In addition, research was conducted to measure respiration rate by analyzing human chest movement in real time. However, the image processing using the infrared thermal image may be difficult to detect the respiratory organ due to the external environmental factors (temperature change, noise, etc.), and thus the accuracy of the measurement of the respiration rate is low.In this study, the images were acquired using visible light and infrared thermal camera to enhance the area of the respiratory tract. Then, based on the two images, features of the respiratory tract region are extracted through processes such as face recognition and image matching. The pattern of the respiratory signal is classified through the k-nearest neighbor classifier, which is one of the statistical classification methods. The respiration rate was calculated according to the characteristics of the classified patterns and the possibility of breathing rate measurement was verified by analyzing the measured respiration rate with the actual respiration rate.

• Journal of KIISE:Computing Practices and Letters
• /
• v.16 no.1
• /
• pp.23-34
• /
• 2010

As the number of user increases according to the improvement of the network, the multi-thread schemes are used to process the service requests of several users who are connected simultaneously. The static thread pool scheme has the problem of occupying a static amount of system resources. On the other hand, the dynamic thread pool scheme can control the number of threads according to the users' requests. However, it has disadvantage that this scheme cannot react to the requests which are larger than the maximum value assigned. In this paper, a web server using a learning-based dynamic thread pool scheme is suggested, which will be running on a server programming of a multi-thread environment. The suggested scheme adds the creation of the threads through the prediction of the next number of periodic requests using Auto Regressive scheme with the web server apache worker MPM (Multi-processing Module). Unlike previous schemes, in order to set the exact number of the necessary threads during the unchanged number of work requests in a certain period, K-Nearest Neighbor algorithm is used to learn the number of threads in advance according to the number of requests. The required number of threads is set by comparing with the previously learned objects. Then, the similar objects are selected to decide the number of the threads according to the request, and they create the threads. In this paper, the response time has decreased by modifying the number of threads dynamically, and the system resources can be used more efficiently by managing the number of threads according to the requests.

• The Journal of Korea Institute of Information, Electronics, and Communication Technology
• /
• v.16 no.6
• /
• pp.489-498
• /
• 2023

Advancements in hardware performance and computing technology have facilitated the progress of climate prediction models to address climate change. The Korea Meteorological Administration employs the GloSea6 model with supercomputer technology for operational use. Various universities and research institutions utilize the Low-GloSea6 model, a low-resolution coupled model, on small to medium-scale servers for weather research. This paper presents an analysis using Intel VTune Profiler on Low-GloSea6 to facilitate smooth weather research on small to medium-scale servers. The tri_sor_dp_dp function of the atmospheric model, taking 1125.987 seconds of CPU time, is identified as a hotspot. Nonlinear regression models, a machine learning technique, are applied and compared to existing functions conducting numerical operations. The K-Nearest Neighbors regression model exhibits superior performance with MAE of 1.3637e-08 and SMAPE of 123.2707%. Additionally, the Light Gradient Boosting Machine regression model demonstrates the best performance with an RMSE of 2.8453e-08. Therefore, it is confirmed that applying a nonlinear regression model to the tri_sor_dp_dp function during the execution of Low-GloSea6 could be a viable alternative.

• Journal of the Institute of Electronics Engineers of Korea CI
• /
• v.49 no.2
• /
• pp.53-60
• /
• 2012

Nearest-neighbor classification predicts the class of an input data with the most frequent class among the near training data of the input data. Even though nearest-neighbor classification doesn't have a training stage, all of the training data are necessary in a predictive stage and the generalization performance depends on the quality of training data. Therefore, as the training data size increase, a nearest-neighbor classification requires the large amount of memory and the large computation time in prediction. In this paper, we propose a prototype selection algorithm that predicts the class of test data with the new set of prototypes which are near-boundary training data. Based on Tomek links and distance metric, the proposed algorithm selects boundary data and decides whether the selected data is added to the set of prototypes by considering classes and distance relationships. In the experiments, the number of prototypes is much smaller than the size of original training data and we takes advantages of storage reduction and fast prediction in a nearest-neighbor classification.

• Korean Journal of Cognitive Science
• /
• v.25 no.2
• /
• pp.91-107
• /
• 2014

The syllable frequency effect is defined as the inhibitory effect that words starting with high frequency syllable generate a longer lexical decision latency and a larger error rate than words starting with low frequency syllable do. Researchers agree that the reason of the inhibitory effect is the interference from syllable neighbors sharing a target's first syllable at the lexical level and the degree of the interference effect correlates with the number of syllable neighbors or stronger syllable neighbors which have a higher word frequency. However, although the syllable frequency can be classified as the syllable type and token frequency, previous studies in visual word recognition have used the syllable frequency without the classification. Recently Conrad, Carreiras, & Jacobs (2008) demonstrated that the syllable type frequency might reflect a sub-lexical processing level including matching from letters to syllables and the syllable token frequency might reflect competitions between a target and higher frequency words of syllable neighbors in the whole word lexical processing level. Therefore, the present study investigated their proposals using word naming tasks. Generally word naming tasks are more sensitive to sub-lexical processing. Thus, the present study expected a facilitative effect of high syllable type frequency and a null effect of high syllable token frequency. In Experiment 1, words starting with high syllable type frequency generated a faster naming latency than words starting with low syllable type frequency with holding syllable token frequency of them. In Experiment 2, high syllable token frequency also created a shorter naming time than low syllable token frequency with holding their syllable type frequency. For that reason, we rejected the propose of Conrad et al. and suggested that both type and token syllable frequency could relate to the sub-lexical processing.