• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.38 no.1
• /
• pp.97-103
• /
• 2001

Hidden Markov Model (HMM) is the most widely used method in speech recognition. In general, HMM parameters are trained to have maximum likelihood (ML) for training data. Although the ML method has good performance, it dose not take account into discrimination to other words. To complement this problem, a word verification method by re-recognition of the recognized word and its similar word using the discriminative function of the two words. To find the similar word, the probability of other words to the HMM is calculated and the word showing the highest probability is selected as the similar word of the mode. To achieve discrimination to each word the weight to each state is appended to the HMM parameter. The weight is calculated by genetic algorithm. The verificator complemented discrimination of each word and reduced the error occurred by similar word. As a result of verification the total error is reduced by about 22%

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.21 no.10
• /
• pp.1886-1891
• /
• 2017

Denoising which is one of major research topics in the image processing deals with recovering the noisy images. Natural images are well known not only for their local but also nonlocal similarity. Patterns of unique edges and texture which are crucial for understanding the image are repeated over the nonlocal region. In this paper, a nonlocal similarity based denoising algorithm is proposed. First for every blocks of the noisy image, nonlocal similar blocks are gathered to construct a overcomplete data set which are inherently sparse in the transform domain due to the characteristics of the images. Then, the sparse transform coefficients are filtered to suppress the non-sparse additive noise. Finally, the image is recovered by aggregating the overcomplete estimates of each pixel. Performance experiments with several images show that the proposed algorithm outperforms the conventional methods in removing the additive Gaussian noise effectively while preserving the image details.

• Proceedings of the Korean Information Science Society Conference
• /
• 1998.10c
• /
• pp.321-323
• /
• 1998

이 논문에서는 크기가 큰 III-Conditioned Matrices 정방행렬의 좌측 또는 우측 역행렬 계산시 계산상의 정확도를 향상시키는 알고리듬에 대하여 기술한다. 이 알고리듬은 대상 행렬의 행벡터들을 Input으로 하고 해당 Input 벡터가 몇번째 행 벡터인지를 나타내는 단위 벡터를 Target 벡터로 하며 초기 Weight 값으로 Pivoting을 겸한 Gauss소거법을 적용하여 얻은 역행렬을 사용하는 Single Layer 인공신경망에 적용하는 역전파 알고리듬과 흡사한 것이다. 각각의 Input 행 벡터에 대하여 역행렬의 열 벡터들이 점진적으로 직교가 되거나 평행이 되도록 근접시키므로써 모든 Input 행 벡터들이 열벡터들에 비교적 균일하게 직교 또는 평행이 되도록 학습시키는 알고리듬이다.

• Proceedings of the Korean Nuclear Society Conference
• /
• 1998.05a
• /
• pp.303-308
• /
• 1998

SMART 계측제어계통 측정신호의 신뢰성을 높이기 위한 실시간 신호검증알고리듬을 개발하였다. 개발된 알고리듬은 선행고장검출행렬, 아날로그 신호용 다중성 기법, 접촉신호용 논리표 알고리듬, 주파수 신호용 다중성 기법 그리고 아날로그 센서 경증을 위한 통계적 모듈의 5개 모듈로 구성되어 있다. 선행고장검출행렬은 측정 신호 중에서 고장의 가능성이 있는 신호를 추출하여 선정된 신호만을 적절한 알고리듬으로 검증하도록 함으로써 전체적인 수행시간을 감소시킨다. 아날로그 신호검증 모듈은 아날로그 측정신호에 대한 물리적/해석적 다중성에 입각하여 고장신호의 크기, 위치를 검출하며, 접촉신호 검증 모듈은 접촉신호들간의 논리값을 비교하여 발생 불가능한 논리값을 가지는 신호를 고장신호로 검출한다. 주파수신호는 아날로그 신호와 유사한 기법을 구현하였으며, 통계적 모듈은 아날로그 센서 자체의 물리적 건전성을 검사하는 모듈이다. 현재 SMART의 설계가 확정되어 있지 않으므로 개발된 신호검증알고리듬을 시험하기 위해서 여러 주요 공정변수가 표현되는 상용 원자로의 냉각재계통을 대상으로 검증 알고리듬을 구현하였으며, 운전모사기로 모사된 신호를 이용하여 개발된 신호검증알고리듬을 시험하였다. 시험결과 각 모듈별로 적절히 고장을 검출함을 보였다.

• Journal of the Institute of Electronics and Information Engineers
• /
• v.49 no.9
• /
• pp.172-180
• /
• 2012

In multi-static sonar for detecting an underwater target, received signals contain the target echo, reverberation and clutter. Clutter and reverberation are main causes of increasing the false alarm rate. MRAL classifies received signals according to the spatial similarity, and it regards classified signal as reflected signals from a reflector. MRAL reduces the false alarm rate this way. However, the results of MRAL can have localization errors. In this paper, an MRAL post processing algorithm is proposed to reduce the localization errors with the least square (LS) method.

• Korean Journal of Remote Sensing
• /
• v.18 no.2
• /
• pp.107-116
• /
• 2002

A hybrid image segmentation algorithm is proposed which integrates edge-based and region-based techniques through the watershed algorithm. First, by using mean curvature diffusion coupled to min/max flow, noise is eliminated and thin edges are preserved. After images are segmented by watershed algorithm, the segmented regions are combined with neighbor regions. Region adjacency graph (RAG) is employed to analyze the relationship among the segmented regions. The graph nodes and edge costs in RAG correspond to segmented regions and dissimilarities between two adjacent regions respectively. After the most similar pair of regions is determined by searching minimum cost RAG edge, regions are merged and the RAG is updated. The proposed method efficiently reduces noise and provides one-pixel wide, closed contours.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2018.11a
• /
• pp.104-107
• /
• 2018

최근 촬영 기기의 기술발전으로 인해 디지털 영상의 해상도가 증가함에 따라 선명한 디지털 영상에 대한 요구가 증가하고 있다. 이러한 요구에도 불구하고 디지털 영상 내 가우시안 잡음 (gaussian noise)은 촬영기기를 통해 영상 획득 및 처리 과정에서 발생하여 화질을 열화 시킨다. 디지털 이미지에서 발생하는 가우시안 잡음을 제거하기 위해서 기존의 저대역 통과 필터 (low-pass filter: LPF)를 사용하면 잡음은 제거되지만, 블러링 현상 (blurring phenomenon)이 나타난다. 이러한 문제점을 개선하기 위해 소벨 연산자 (sobel operator)를 사용하여 영상 내 에지 맵 (edge-map)을 생성하여 에지 영역과 동질 영역을 구분한다. 에지영역에서는 약한 저역 필터 (weak low-pass filter)를 사용하고, 그 외의 이미지 영역에서는 강한 저역 필터 (strong low-pass filter)를 사용하는 알고리듬을 제안하였다. 그리고 다양한 이미지에 대하여 기존 알고리듬과 제안한 알고리듬의 적용한 결과를 통해 주관적 화질 비교하였고 객관적 지표로 최대 신호 대 잡음비 (peak signal-to noise ratio: PSNR)와 구조 유사성 (structural similarity: SSIM)을 사용하여 성능을 평가하였다. 실험결과를 통해 제안된 알고리듬이 잡음 제거 및 외곽선 보존의 우수함을 확인하였다.

• Journal of the Korea Computer Industry Society
• /
• v.2 no.8
• /
• pp.1123-1132
• /
• 2001

We show that image make converts of multiplex resolution using wavelet transform algorithm. To evaluate the robustness, we have measured comparison of similarity using various sequences which is inserted important coefficient of middle frequency band. The wavelet transform is advantage that it has a special quality of frequency domain and a special quality of spatial domain. As result of experiments, it is to certify that The chaos sequence similarity is higher than other sequence. So the chaos sequence will be used for watermark sequence.

• Journal of the Korea Society of Computer and Information
• /
• v.6 no.4
• /
• pp.21-29
• /
• 2001

We show that image make converts of multiplex resolution using wavelet transform algorithm. To evaluate the robustness. we have measured comparison or similarity using various sequences which is inserted important coefficient of middle frequency band. The wavelet transform is advantage that it has a special quality of frequency domain and a special quality of spatial domain. Watermark is used pseudo random number, gaussian sequence, chaos sequence and sobel sequence. As result of experiments, it is to certify that The chaos sequence similarity is higher than other sequence. So the chaos sequence will be used for watermark sequence.

• Journal of IKEEE
• /
• v.23 no.3
• /
• pp.910-916
• /
• 2019

Non-local similarity of natural images is one of highly exploited features in various applications dealing with images. Unique edges, texture, and pattern of the images are frequently repeated over the entire image. Once the similar image blocks are classified into a cluster, representative features of the image blocks can be extracted from the cluster. The bigger the size of the cluster is the better the additive white noise can be separated. Denoising is one of major research topics in the image processing field suppressing the additive noise. In this paper, a denoising algorithm is proposed which first clusters the noisy image blocks based on similarity, extracts the feature of the cluster, and finally recovers the original image. Performance experiments with several images under various noise strengths show that the proposed algorithm recovers the details of the image such as edges, texture, and patterns while outperforming the previous methods in terms of PSNR in removing the additive Gaussian noise.