• Proceedings of the Korean Information Science Society Conference
• /
• 2003.10b
• /
• pp.595-597
• /
• 2003

오프라인 필기체 숫자 인식에서 다양한 변형을 잘 흡수 할 수 있는 효율적인 특징을 찾는 것은 중요한 일이며, 본 논문에서는 이를 위해 다양한 단일특징들을 구현 하였으며, 단일 특징만으로는 만족 할 만한 성능을 기대하기 어렵기 때문에 다양한 단일 특징을 복합특징으로 구성하였다. 또한 오프라인 필기체 숫자인식에서 좋은 성능을 발휘하는 것으로 알려진 신경회로망으로 학습을 하였으며, 인식의 성능을 개선시키기 위해 효과적인 특징을 조합하여 하나의 단일 신경회로망들을 구성하고 그것을 다시 복합신경회로망으로 구성하여 성능을 실험 함으로서 성능의 향상을 볼 수 있었고, 신경회로망에 더하여 성능을 개선시키기 위해 신경회로망을 보완 할 수 있는 다수결 투표 방법을 사용하였다. 본 논문에서는 신경회로망의 인식 결과를 비교 분석하여 최적의 특징을 찾아 낸 결과를 2차 다수결 투표를 사용하여 인식하는 방법을 제안한다. 제안된 방식의 성능을 검증하기 위해서 Concorida 대학교의 CENPARIMI 숫자 데이터 베이스를 가지고 인식을 수행 하였으며. 그 결과 97.40%의 정인식률과 0.75%의 오인식률 그리고 1.85%의 거부률을 보였다.

• The Journal of the Acoustical Society of Korea
• /
• v.25 no.6
• /
• pp.277-281
• /
• 2006

This paper presents a method for the automatic recognition of pitch accents over syllables. The method that we propose is based on the time-delay recursive neural network (TDRNN). which is a neural network classifier with two different representation of dynamic context: the delayed input nodes allow the representation of an explicit trajectory F0(t) along time. while the recursive nodes provide long-term context information that reflects the characteristics of pitch accentuation in spoken English. We apply the TDRNN to pitch accent recognition in two forms: in the normal TDRNN. all of the prosodic features (pitch. energy, duration) are used as an entire set in a single TDRNN. while in the distributed TDRNN. the network consists of several TDRNNs each taking a single prosodic feature as the input. The final output of the distributed TDRNN is weighted sum of the output of individual TDRNN. We used the Boston Radio News Corpus (BRNC) for the experiments on the speaker-independent pitch accent recognition. π 1e experimental results show that the distributed TDRNN exhibits an average recognition accuracy of 83.64% over both pitch events and non-events.

• The Journal of the Korea Contents Association
• /
• v.1 no.1
• /
• pp.8-14
• /
• 2001

In this paper, we suggests a algorithms of recognizing the disease region by extracting particular organ from medical image. This method can extract liver region in spite of input image including many organs and charged format by using multi-threshold of feed-back-structure for segmentation liver region, and suggest the recognition of disease region in extracted liver, using multi-neural network structured by RBF and BP, overcoming the defect of single-neural network. The algorithm in this paper is proficient in adaptation for a multi form change of input medical image. This algorithm can be used at tole-medicine through automatic recognition after recognizing of the disease region by real-tire medical Image.

• Journal of the Institute of Electronics Engineers of Korea SD
• /
• v.45 no.6
• /
• pp.16-21
• /
• 2008

This paper introduces a new type of Hopfield neural network using newly developed single-electron devices. In the electrical model of the Hopfield neural network, a single-electron synapse, used as a voltage(or current)-variable resistor, and two stages of single-electron inverters, used as a nonlinear activation function, are simulated with a single-electron circuit simulator using Monte-Carlo method to verily their operation.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.11 no.4
• /
• pp.325-330
• /
• 2001

It has been realized that the results of circuit level simulation of neural networks, used for optimization problems, arc much different from those of algorism level simulation. In other words, the outputs converges asymptotically as time elapes, however, the input convergence depends on the value of parasitic conductance connected between input node and ground. Also, this conductance affects system performance. This paper discusses the influence of input conductance on the convergece of the continuous Hopfield neural networks. The convergence has been analyzed for the input and output nodes of neurons. Also, the characteristics of equilibrium points has been analyzed depending on different values of the input conductance.

• Journal of the Korea Convergence Society
• /
• v.9 no.6
• /
• pp.39-44
• /
• 2018

The application of deep neural networks to finance has received a great deal of attention from researchers because no assumption about a suitable mathematical model has to be made prior to forecasting and they are capable of extracting useful information from large sets of data, which is required to describe nonlinear input-output relations of financial time series. The paper presents a new deep neural network model where single layered autoencoder and 4 layered neural network are serially coupled for stock price forecasting. The autoencoder extracts deep features, which are fed into multi-layer neural networks to predict the next day's stock closing prices. The proposed deep neural network is progressively learned layer by layer ahead of the final learning of the total network. The proposed model to predict daily close prices of KOrea composite Stock Price Index (KOSPI) is built, and its performance is demonstrated.

• Journal of Korea Society of Industrial Information Systems
• /
• v.25 no.4
• /
• pp.29-38
• /
• 2020

In this research, we proposed the active pulse classification algorithm using multi-label convolutional neural networks for active sonar system. The proposed algorithm has the advantage of being able to acquire the information of the active pulse at a time, unlike the existing single label-based algorithm, which has several neural network structures, and also has an advantage of simplifying the learning process. In order to verify the proposed algorithm, the neural network was trained using sea experimental data. As a result of the analysis, it was confirmed that the proposed algorithm converged, and through the analysis of the confusion matrix, it was confirmed that it has excellent active pulse classification performance.

• Journal of the Korean Society for Nondestructive Testing
• /
• v.30 no.4
• /
• pp.302-310
• /
• 2010

In this paper, we studied Bagging neural network for predicting defect size of steam generator(SG) tube in nuclear power plant. Bagging is a method for creating an ensemble of estimator based on bootstrap sampling. For predicting defect size of SG tube, we first generated eddy current testing signals for 4 defect patterns of SG tube with various widths and depths. Then, we constructed single neural network(SNN) and Bagging neural network(BNN) to estimate width and depth of each defect. The estimation performance of SNN and BNN were measured by means of peak error. According to our experiment result, average peak error of SNN and BNN for estimating defect depth were 0.117 and 0.089mm, respectively. Also, in the case of estimating defect width, average peak error of SNN and BNN were 0.494 and 0.306mm, respectively. This shows that the estimation performance of BNN is superior to that of SNN.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2001.05a
• /
• pp.456-459
• /
• 2001

In this paper, a wavelet neural network for adaptive control was proposed. The structure of this network is similar to that of the multilayer perceptron(MLP), except that here the sigmoid functions are replated by mother wavelet function in the hidden units. The simulation result showed the effectiveness of using the wavelet neural network structure in the adaptive control of one-link manipulator.

• Proceedings of the KIEE Conference
• /
• 1999.07g
• /
• pp.2812-2814
• /
• 1999

그레이 영상 입출력이 가능하고, 다양한 영상 크기에 적용 가능한 아날로그 셀룰라 신경회로망을 설계하였다. 아날로그 셀룰라 신경회로망은 실시간 병렬처리가 가능하므로, 영상처리 패턴인식과 같은 분야에 유용하게 사용될 수 있다. 기존의 하드웨어로 구현된 셀를라 신경회로망은 이진 영상를 출력하고, 단일 칩에 구현할 수 있는 셀의 수에 제한이 있기 때문에 범용의 영상처리에 응용하기에 적합지 않다. 본 연구에서 설계된 셀룰라 신경회로망은 영상 입력 크기의 분해능을 향상시켜 그레이 영상 처리가 가능한 칩을 설계하였다. 설계된 셀룰라 신경회로망를 이용한 그레이 영상의 에지추출 시뮬레이션 결과, 선명한 에지 영상을 얻을 수 있었다