• Proceedings of the Korean Institute of Electrical and Electronic Material Engineers Conference
• /
• 2003.11a
• /
• pp.558-562
• /
• 2003

In this paper, neural networks is studied to apply as a PD source classification in XLPE power cable specimen. For treeing discharge sources in the specimen, three defected models are made. And these data making use of a computer-aided discharge analyser, statistical and other discharge parameters is calculated to discrimination between different models of discharge sources. And also these parameter is applied to classify PD sources by neural networks. Neural Networks has good recognition rate for three PD sources.

• Proceedings of the KSPS conference
• /
• 2005.11a
• /
• pp.155-158
• /
• 2005

Keyword spotting is effective to find keyword from the continuously pronounced speech. However, non-keyword may be accepted as keyword when the environmental noise occurs or speaker changes. To overcome this performance degradation, utterance rejection techniques using confidence measure on the recognition result have been developed. In this paper, we apply DTW to the HMM based broadcasting news keyword spotting system for rejecting non-keyword. Experimental result shows that false acceptance rate is decreased to 50%.

• Proceedings of the KSPS conference
• /
• 2003.05a
• /
• pp.203-206
• /
• 2003

In this paper, we employ the adaptive comb filtering for effective noise reduction in mobile communication environment. Adaptive comb filtering is a well- known method for noise reduction, but requires the correct pitch period and must be applied just in voiced speech frames. To satisfy these requirements we use two kinds of information extracted from speech packets, one of which is the pitch period information measured precisely by a speech coder and the other is the frame rate information related to a decision on speech or silence frame. Experiments on speech recognition system confirm the efficiency of this method. Feature parameters employing this method give superior performance in noise environment to those extracted directly from output speech.

• Proceedings of the Korea Institute of Fire Science and Engineering Conference
• /
• 1997.11a
• /
• pp.282-288
• /
• 1997

The purpose of this paper is to predict the situation of leak in closed space using an Artificial Neural Network (ANN). The existing system can't monitor the whole He situations with on/off signals. Especially the first stage of data determines the leak spot and intensity is disregarded in gas accidents. To complement these faults, a new prototype of monitoring system is proposed. Ihe system is composed of'sensing systenL data acquisition system computer, and ANN implemented in software and is capable of identifying the leak spot and intensity in closed space. The concentration of gas is measured at the 4 different places. The network has 3 layers that are composed of 4 input Processing Element (PE),24 hidden PEs, md 4 output PEs. The ANN has optimum condition through several experiments and as a consequence the recognition rate of93.75% is achieved finally

• Journal of Korea Multimedia Society
• /
• v.18 no.1
• /
• pp.71-80
• /
• 2015

Many researches have been studied to overcome the weak points in authentication schemes of mobile devices such as pattern-authentication that is vulnerable for smudge-attack. Since random-pattern-lock authenticates users by drawing figure of predefined-shape, it can be a method for robust security. However, the authentication performance of random-pattern-lock is influenced by input noise and individual characteristics sign pattern. We introduce an optimization method of user input direction to increase the authentication accuracy of random-pattern-lock. The method uses the likelihood of each direction given an data which is angles of line drawing by user. We adjusted recognition range for each direction and achieved the authentication rate of 95.60%.

• Journal of Korea Multimedia Society
• /
• v.18 no.1
• /
• pp.1-8
• /
• 2015

In recent years, ear has emerged as a new biometric trait, because it has advantage of higher user acceptance than fingerprint and can be captured at remote distance in an indoor or outdoor environment. This paper proposes an individual identification method using ear region based on SIFT(shift invariant feature transform). Unlike most of the previous studies using rectangle shape for extracting a region of interest(ROI), this study sets an ROI as a flexible expanded region including ear. It also presents an effective extraction and matching method for SIFT keypoints. Experiments for evaluating the performance of the proposed method were performed on IITD public database. It showed correct identification rate of 98.89%, and it showed 98.44% with a deformed dataset of 20% occlusion. These results show that the proposed method is effective in ear recognition and robust to occlusion.

• Journal of the Korea Safety Management & Science
• /
• v.5 no.4
• /
• pp.1-12
• /
• 2003

This study conducted the questionnaire on present condition of certification, background, purpose, effect and promotion of an occupational health & safety management system which introduced in domestic from 1999'. The questionnaire was conducted with 136 companies which replied in questionnaire distributed to 193 companies which received the certification. Also, recognition improvement of managers and participation of workers are demanded, the support to subcontract companies is demanded because the disaster prevention effect of the companies which possess the support program on subcontract companies is good, and reduction in the insurance rate to companies which received certification is demanded to activate an occupational health & safety management system.

• Korean Journal of Optics and Photonics
• /
• v.3 no.3
• /
• pp.191-197
• /
• 1992

We have developed a new optical associative memory system hased on the symmetric three layered neural network model, uhing two holograms and a LCIV. In the experiment, four Korean alphabet letters (ㄹ, ㅅ, ㅇ, ㅈ) are used as memory patterns. The results are compared with those of the two layered network and the IIopfield models. The results show that more than 95% recognition ablity is obtained for thc input which has the error rate less than 12%.

• Electronics and Telecommunications Trends
• /
• v.34 no.1
• /
• pp.98-110
• /
• 2019

Machine learning has been applied to medical imaging and has shown an excellent recognition rate. Recently, there has been much interest in preventive medicine. If data are accessible, machine learning packages can be used easily in digital healthcare fields. However, it is necessary to prepare the data in advance, and model evaluation and tuning are required to construct a reliable model. On average, these processes take more than 80% of the total effort required. In this study, we describe the basic concepts of machine learning, pre-processing and visualization of datasets, feature engineering for reliable models, model evaluation and tuning, and the latest trends in popular machine learning frameworks. Finally, we survey a explainable machine learning analysis tool and will discuss the future direction of machine learning.

• Journal of the Semiconductor & Display Technology
• /
• v.19 no.4
• /
• pp.65-70
• /
• 2020

CNN's technologies that represent emotional detection include primitive CNN algorithms, deployment normalization, and drop-off. We present the methods and data of the three experiments in this paper. The training database and the test database are set up differently. The first experiment is to extract emotions using Batch Normalization, which complemented the shortcomings of distribution. The second experiment is to extract emotions using Dropout, which is used for rapid computation. The third experiment uses CNN using convolution and maxpooling. All three results show a low detection rate, To supplement these problems, We will develop a deep learning algorithm using feature extraction method specialized in image processing field.