• Proceedings of the Korean Information Science Society Conference
• /
• 2005.07b
• /
• pp.868-870
• /
• 2005

본 논문은 포즈가 취해진 얼굴의 정확한 특징점 추출을 위하여 적응적인 평균 모양 방법을 이용한 ASM(Active Shape Model)을 제안한다. ASM은 사람 얼굴의 모양을 모델링하기 위하여 통계학상의 모양 모델을 이용한다. 통계학상의 모양 모델의 평균 모양은 입력 영상의 얼굴 포즈와 관계없이 하나로 고정되어 있으며, 이는 모양 모델 제한 조건 검사 및 복원과정에서 잘못된 결과를 만드는 원인이 된다. 이러한 문제를 해결하기 위하여 입력 영상의 얼굴 모양에 적응적인 평균 모양을 제안하며, 실험을 통해 제안한 방법이 고정된 평균 모양 방법의 문제를 해결하고 특징점 추출 성능을 향상시킴을 보였다.

• Proceedings of the KIEE Conference
• /
• 2007.07a
• /
• pp.321-322
• /
• 2007

본 논문에서는 로봇과 인간의 인터액션을 위해 인간 몸의 실루엣 및 액티브 스켈레톤 추출 기법을 제안한다. 연속된 이미지 정보로 부터 얻어진 옷영역등의 정보에서 background subtraction를 이용한 adaptive fusion을 통해 추출된 인간 몸의 실루엣을 바탕으로 active contour와 가상 신체 모델인 skeleton model을 응용하여 작은 움직임에 보다 강한 active skeleton model을 이용하여 인간 몸의 특징 점 위치를 추출하는 방법을 한다.

• Proceedings of the KAIS Fall Conference
• /
• 2010.05a
• /
• pp.249-252
• /
• 2010

LPC 계수를 LSP 변환하는 방법에는 복소근, 실근, 비율 필터, 체비셰프 급수, 적응적 순차형 최소제곱 평균 방법(adaptive sequential LMS) 등이 있다. 이 방법들 중 음성 부호화기에서 주로 사용하는 실근 방법은 근을 구하기 위해 주파수 영역을 순차적으로 검색하기 때문에 계산시간이 많이 소요되는 단점을 갖는다. 본 논문에서는 LPC에서 LSP로 변환하는 4가지 고속 알고리즘을 제안한다. 첫 번째 방식에서는 검색간격에 멜 스케일을 적용하였고, 두 번째는 홀수번째 LSP 파라미터의 분포도를 이용하여 검색순서를 조정한 방법이다. 세 번째 방식과 네 번째 방식에서는 각각, 모음 특성, LSP 분포특성과 해상도를 이용하여 계산시간을 단축하였다. LSP 변환시간은 4가지 방법 모두 35~50% 단축되었다. 또한 실험결과에서는 각 알고리즘의 고유한 특성에 대하여 분석한다.

• Proceedings of the Optical Society of Korea Conference
• /
• 2000.08a
• /
• pp.46-47
• /
• 2000

본 논문에서는 적응광학시스템$^{(1)}$ 의 성능 향상에 필수적인 파면왜곡의 기울기 정보를 고속으로 측정하기 위한 중심점 추출 알고리즘을 제안하였다. 본 논문에서는 컴퓨터 내부의 영상처리전용보드와 CCD카메라를 이용하여 하트만 센싱 점 영상을 획득하였고, 획득한 하트만 센싱 점 영상에 대해 제안한 중심점 추출 알고리즘을 적용하여 서브픽셀 분해능으로 X축과 Y축의 기울기 정보를 고속으로 추출하였다. CCD센서에 촬상되는 하트만 센싱 점영상에서 각각의 점 영상은 중심점으로부터 대칭형으로 강도가 분포되어 있다고 가정할 수 있으나 전체 점영상의 각 점을 분석한 결과 비대칭적으로 예외적인 강도 분포를 갖는 점영상도 일부 발견되었다. 파면 왜곡이 없는 하트만 센싱 점영상으로부터 X, Y축 파면 왜곡 기울기 값을 추출한 결과 CCD 센서 픽셀의 기저 노이즈가 큰 불안정한 영역에서 기울기 값이 반복적으로 크게 추출되어 파면왜곡보정 시스템의 보정 성능을 떨어뜨리는 효과가 나타났다. (중략)

• Journal of the Korean Society for Industrial and Applied Mathematics
• /
• v.27 no.1
• /
• pp.37-55
• /
• 2023

In this paper we propose a new algorithm for detecting and counting flowers in a complex background based on digital images. The algorithm mainly includes the following parts: edge contour extraction of flowers, edge contour determination of overlapped flowers and flower counting. We use a contour detection technique in Computer Vision (CV) to extract the edge contours of flowers and propose an improved algorithm with a concave point detection technique to find accurate segmentation for overlapped flowers. In this process, we first use the polygon approximation to smooth edge contours and then adopt the second-order central moments to fit ellipse contours to determine whether edge contours overlap. To obtain accurate segmentation points, we calculate the curvature of each pixel point on the edge contours with an improved Curvature Scale Space (CSS) corner detector. Finally, we successively give three adaptive judgment criteria to detect and count flowers accurately and automatically. Both experimental results and the proposed evaluation indicators reveal that the proposed algorithm is more efficient for flower counting.

• Journal of the Korea Computer Industry Society
• /
• v.3 no.9
• /
• pp.1145-1156
• /
• 2002

A method to measure the speed and distance of moving object is proposed using the stereo vision system. One of the most important factors for measuring the speed and distance of moving object is the accuracy of object tracking. Accordingly, the background image algorithm is adopted to track the rapidly moving object and the local opening operator algorithm is used to remove the shadow and noise of object. The extraction efficiency of moving object is improved by using the adaptive threshold algorithm independent to variation of brightness. Since the left and right central points are compensated, the more exact speed and distance of object can be measured. Using the background image algorithm and local opening operator algorithm, the computational processes are reduced and it is possible to achieve the real-time processing of the speed and distance of moving object. The simulation results show that background image algorithm can track the moving object more rapidly than any other algorithm. The application of adaptive threshold algorithm improved the extraction efficiency of the target by reducing the candidate areas. Since the central point of the target is compensated by using the binocular parallax, the error of measurement for the speed and distance of moving object is reduced. The error rate of measurement for the distance from the stereo camera to moving object and for the speed of moving object are 2.68% and 3.32%, respectively.

• The Journal of the Acoustical Society of Korea
• /
• v.41 no.3
• /
• pp.326-334
• /
• 2022

In this paper, we propose speech enhancement algorithm as a pre-processing for robust speech recognition in noisy environments. Auxiliary-function-based Independent Vector Analysis (AuxIVA) is performed with weighted covariance matrix using time-varying variances with scaling factor from target masks representing time-frequency contributions of target speech. The mask estimates can be obtained using Neural Network (NN) pre-trained for speech extraction or diffuseness using Coherence-to-Diffuse power Ratio (CDR) to find the direct sounds component of a target speech. In addition, outputs for omni-directional noise are closely chained by sharing the time-varying variances similarly to independent subspace analysis or IVA. The speech extraction method based on AuxIVA is also performed in Independent Low-Rank Matrix Analysis (ILRMA) framework by extending the Non-negative Matrix Factorization (NMF) for noise outputs to Non-negative Tensor Factorization (NTF) to maintain the inter-channel dependency in noise output channels. Experimental results on the CHiME-4 datasets demonstrate the effectiveness of the presented algorithms.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.10 no.12
• /
• pp.2283-2288
• /
• 2006

As development in technology of bioinformatics recently makes it possible to operate micro-level experiments, we can observe the expression pattern of total genome through on chip and analyze the interactions of thousands of genes at the same time. Thus, DNA microarray technology presents the new directions of understandings for complex organisms. Therefore, it is required how to analyze the enormous gene information obtained through this technology effectively. In this thesis, We used sample data of bioinformatics core group in harvard university. It designed and implemented system that evaluate accuracy after dividing in class of two using Bayesian algorithm, ASA, of feature extraction method through normalization process, reducing or removing of noise that occupy by various factor in microarray experiment. It was represented accuracy of 98.23% after Lowess normalization.

• Journal of KIISE:Software and Applications
• /
• v.27 no.3
• /
• pp.248-256
• /
• 2000

This paper presents a personalized web agent that constructs user profile which consists of user preferences on the web and recommends his/her relevant information to the user. The personalized web agent consists of monitor agent, user profile construction agent, and user profile refinement agent. The monitor agent makes a user describe his/her preferences directly and it creates the database of preference document, finally performs several keyword extraction to increase the accuracy of the DB. The user profile construction agent transforms the extracted keywords into user profile that could be confirmed and edited by the user. and the refinement agent refines user profile by recursively learning and processing user feedback. In this paper, we describe the several keyword weighting and inductive learning techniques in detail. Finally, we describe the adaptive web retrieval and push agent that perform adaptive services to the user.

• The Journal of the Korea Contents Association
• /
• v.7 no.2
• /
• pp.40-50
• /
• 2007

In this paper, we propose an efficient method for detecting and tracking multiple moving objects based on morphological region merging from real-time video sequences. The proposed approach consists of adaptive threshold extraction, morphological region merging and detecting and tracking of objects. Firstly, input frame is separated into moving regions and static regions using the difference of images between two consecutive frames. Secondly, objects are segmented with a reference background image and adaptive threshold values, then, the segmentation result is refined by morphological region merge algorithm. Lastly, each object segmented in a previous step is assigned a consistent identification over time, based on its spatio-temporal information. The experimental results show that a proposed method is efficient and useful in terms of real-time multiple objects detecting and tracking.