• Proceedings of the Korean Society of Surveying, Geodesy, Photogrammetry, and Cartography Conference
• /
• 2006.04a
• /
• pp.193-198
• /
• 2006

Among 242 Hyperion bands, there are 46 bands that contain completely no information and some other bands with various kinds of noise. It is mainly due to the atmosphenc absorption and the low signal-to-noise ratio. The visual inspection for selecting clean and stable bands is a simple practice, but is a manual, inefficient, and subjective Process. Though uncalibrated, overlapping, and all deep water absorption bands are removed, there still exist noisy bands. In this paper, we propose that the extrema ratio be measured for noise estimation and the unsupervised band selection be performed using the Expectation-Maximization algorithm. The Hyperion data were classified into 5 categories according to the image quality by visual inspection, and used as the reference data. The accuracy of the proposed method was compared with signal-to-noise ranking and entropy ranking. As a result, the proposed mettled was effective as preprocessing step for band selection.

• The Transactions of The Korean Institute of Electrical Engineers
• /
• v.60 no.6
• /
• pp.1183-1189
• /
• 2011

In this study, we propose the unsupervised image segmentation algorithm to estimate dental plaque accumulations on digital imaging with methylene blue disclosed plaque. With this aim, RGB color plane is mapped into HSI coordinates and the circular histogram of Hue is reconstructed by applying Otsu's threshold level. The histogram distribution on Saturation features is also analyzed by maximizing the variance between a plaque candidate and non-plaque one. The dental plaque regions are resolved by applying the composite decision logics based on the threshold level of Hue and Saturation.

• Proceedings of the Korean Operations and Management Science Society Conference
• /
• 1995.04a
• /
• pp.587-599
• /
• 1995

This paper proposes a method for pattern recognition using spectral analysis and a hybrid genetic-neural-fuzzy networks. The feature vectors using spectral analysis on contour sequences of 2-D images are extracted, and the vectors are not effected by translation, rotation and scale variance. A combined model using the advantages of conventional method is proposed, those are supervised learning BP, global searching genetic algorithm, and unsupervised learning fuzzy c-method. The proposed method is applied to 10 aircraft recognition to confirm the performance of the method. The experimental results show that the proposed method is better accuracy than conventional method using BP or fuzzy c-method, and learning speed is enhanced.

• Proceedings of the KIEE Conference
• /
• 2004.07d
• /
• pp.2537-2539
• /
• 2004

SOM은 심전도 신호의 진단에 있어서 효과적인 Clustering을 해주는 신경망이라는 것을 몇몇의 실험을 통하여 알 수 있었다. [1] 하지만 출력노드의 크기를 임의로 지정해야 하는 문제점이 있고 일반적으로 출력층의 크기가 클수록 진단결과는 좋지만 인간시간은 오래걸린다는 단점이 있다. 따라서 진단능력과 학습속도 사이의 균형에 관련된 문제가 대두되게 된다. 본 논문에서는 이러한 문제점을 극복하고자 기존의 SOM 신경망의 단점을 보완하고자 GCS(Growing Cell Structures)를 이용한 심전도의 학습속도와 분류능력 사이의 효율성 개선 방안을 제안한다. 이 방범은 GCS를 이용하여 적절한 노드의 수를 찾아내는 것이다. 이를 이용한 심전도 진단의 실험을 통해 기존의 SOM이 할 수 없었던 자체적인 출력노드의 증감을 행함을 확인할 수 있었다. 또한 출력노드의 감소로 인해 연산량이 줄어 학습시간의 효율성이 증가하였다.

• 제어로봇시스템학회:학술대회논문집
• /
• 2001.10a
• /
• pp.127.1-127
• /
• 2001

In this paper we propose an input/output model based fault diagnosis method to detect and isolate single faults in the robot arm control system. The proposed algorithm is functionally composed of three main parts-parameter estimation, fault detection, and isolation, When a change in the system occurs, the errors between the system output and the estimated output cross a predetermined threshold, and once a fault in the system is detected, and in this zone the estimated parameters are transferred to the fault classifier by ART2(adaptive resonance theory 2) neural network for fault isolation. Since ART2 neural network is an unsupervised neural network fault classifier does not require the knowledge of all possible faults to isolate the faults occurred in the system. Simulations are carried out to evaluate the performance of the proposed ...

• ETRI Journal
• /
• v.39 no.1
• /
• pp.21-29
• /
• 2017

This paper addresses the problem of unsupervised speech separation based on robust non-negative matrix factorization (RNMF) with ${\beta}$-divergence, when neither speech nor noise training data is available beforehand. We propose a robust version of non-negative matrix factorization, inspired by the recently developed sparse and low-rank decomposition, in which the data matrix is decomposed into the sum of a low-rank matrix and a sparse matrix. Efficient multiplicative update rules to minimize the ${\beta}$-divergence-based cost function are derived. A convolutional extension of the proposed algorithm is also proposed, which considers the time dependency of the non-negative noise bases. Experimental speech separation results show that the proposed convolutional RNMF successfully separates the repeating time-varying spectral structures from the magnitude spectrum of the mixture, and does so without any prior training.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.9 no.6
• /
• pp.571-576
• /
• 1999

We consider a class of auto-associative memories namely N-Cubes (Neural-network Cubes) in which 2-D gray-level images and hidden sinusoidal 1-D wavelets are stored in cubical memories. First we develop a learning procedure based upon the least-squares algorithm, Therefore each 2-D training image is mapped into the associated 1-D waveform in the training phase. Second we show how the recall procedure minimizes errors among the orthogonal basis functions in the hidden layer. As a 2-D images ould be retrieved in the recall phase. Simulation results confirm the efficiency and the noise-free properties of N-Cubes.

• Proceedings of the Korea Water Resources Association Conference
• /
• 2007.05a
• /
• pp.1427-1431
• /
• 2007

In this study, the new methodology which combines Kohonen self-organizing map(KSOM) neural networks model and the conventional neural networks models such as feedforward neural networks model and generalized neural networks model is introduced to forecast flood stage in Nakdong river, Republic of Korea. It is possible to train without output data in KSOM neural networks model. KSOM neural networks model is used to classify the input data before it combines with the conventional neural networks model. Four types of models such as SOM-FFNNM-BP, SOM-GRNNM-GA, FFNNM-BP, and GRNNM-GA are used to train and test performances respectively. From the statistical analysis for training and testing performances, SOM-GRNNM-GA shows the best results compared with the other models such as SOM-FFNNM-BP, FFNNM-BP, and GRNNM-GA and FFNNM-BP shows vice-versa. From this study, we can suggest the new methodology to forecast flood stage and construct flood warning system in river basin.

• Proceedings of the IEEK Conference
• /
• 2002.07c
• /
• pp.1674-1677
• /
• 2002

In this paper, the new method to extract the features of iris signals is proposed; Multiresolution ICA (M-ICA) provides good properties to represent signals with time-frequency. The conventional methods were to use the technique of filter bank analysis, while ICA is unsupervised learning algorithm using high-order statistics. M-ICA could make use of strengths of learn- ing method and multiresolution. Also, we performed comparative studies of different feature extraction techniques applied to personal identification using iris pat- tern. To measure goodness of methods, we use Fisher’s discriminant ratio to quantify the class-separability of features generated by various techniques.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2002.11b
• /
• pp.869-872
• /
• 2002

네트워크 기반 침입탐지시스템은 연속적으로 발생하는 패킷의 무손실 축소와, 패킷으로 정상 또는 비정상 행위패턴을 정확히 모델링한 모델 생성이 전체성능을 판단하는 중요한 요소가 된다. 네트워크 기반 비정상행위 판정 침입탐지시스템에서는 이러한 탐지모델 구축을 위해 주로 감독학습 알고리즘을 사용한다. 본 논문은 탐지모델 구축에 사용하는 감독 학습 방식이 가지는 문제점을 지적하고, 그에 대한 대안으로 비감독 학습방식의 학습알고리즘을 제안한다. 감독 학습을 사용하여 탐지모델을 구축하기 위해서는 정상행위의 패킷을 취합해야 하는 사전 부담이 있는 반면에 비감독 학습을 사용하게 되면 이러한 사전작업 없이 탐지모델을 구축할 수 있다. 본 논문에서는 비감독학습 알고리즘을 비교 분석하기 위해서 COBWEB, k-means, Autoclass 알고리즘을 사용했으며, 성능을 평가하기 위해서 비정상행위도(Abnormal Behavior Level)를 계산하여 에러율을 구하였다.