• 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%

• The Korean Journal of Emergency Medical Services
• /
• v.28 no.2
• /
• pp.7-19
• /
• 2024

Purpose: To adjust item numbers in a national test, this study used item response theory to examine changes in average scores, reliability, difficulty, and discrimination according to the adjustment of item numbers. Methods: We analyzed the dichotomous coding of correct and incorrect answers of 473 examinees in a mock test conducted in 2023. Additionally, as an explanatory pilot study, we used an online questionnaire to survey experts on their perceptions of the appropriate item numbers for each test subject from January 18, 2024, to February 15, 2024. Results: Regarding the item numbers on the national exam, experts preferred to reduce the number of management of emergency patients (33.14±6.09, p<.05) and advanced emergency medical care: subtopics (104.49±11.55, p<.05), and the total number of questions (217.82±20.95, p<.05). In a simulation set in which items with low item fit were removed after fitting a two-parameter item response theory model, reliability was maintained at .910 until the 5th test consisting of 185 questions with little loss of difficulty, discrimination, and average score, and there was no correlation between item numbers and average score. Conclusion: Experts responded that reducing the number of items on the national exam was appropriate. As a result of the item reduction simulation, there was no significant loss in the average score, difficulty, discrimination, or reliability. More reliable results could be obtained if the results were based on a validity analysis and analyzed using actual national exams.

• Nuclear Engineering and Technology
• /
• v.56 no.8
• /
• pp.3123-3128
• /
• 2024

The dicentric chromosome assay is a gold standard method to estimate radiation exposure by calculating the ratio of dicentric chromosomes existing in cells. The objective of this study was to propose an automatic dicentric chromosome discrimination method based on deep convolutional neural networks using radiation exposure patient data. From 45 patients with radiation exposure, conventional Giemsa-stained images of 116,258 normal and 2800 dicentric chromosomes were confirmed. ImageNet was used to pre-train VGG19, which was modified and fine-tuned. The proposed modified VGG19 demonstrated dicentric chromosome discrimination performance, with a true positive rate of 0.927, a true negative rate of 0.997, a positive predictive value of 0.882, a negative predictive value of 0.998, and an area under the receiver operating characteristic curve of 0.997.

• Korean Journal of Radiology
• /
• v.21 no.9
• /
• pp.1055-1064
• /
• 2020

Objective: To assess the incremental prognostic value of coronary computed tomography angiography (CCTA) in comparison to a clinical risk model (Framingham risk score, FRS) and coronary artery calcium score (CACS) for future cardiac events in ischemic stroke patients without chest pain. Materials and Methods: This retrospective study included 1418 patients with acute stroke who had no previous cardiac disease and underwent CCTA, including CACS. Stenosis degree and plaque types (high-risk, non-calcified, mixed, or calcified plaques) were assessed as CCTA variables. High-risk plaque was defined when at least two of the following characteristics were observed: low-density plaque, positive remodeling, spotty calcification, or napkin-ring sign. We compared the incremental prognostic value of CCTA for major adverse cardiovascular events (MACE) over CACS and FRS. Results: The prevalence of any plaque and obstructive coronary artery disease (CAD) (stenosis ≥ 50%) were 70.7% and 30.2%, respectively. During the median follow-up period of 48 months, 108 patients (7.6%) experienced MACE. Increasing FRS, CACS, and stenosis degree were positively associated with MACE (all p < 0.05). Patients with high-risk plaque type showed the highest incidence of MACE, followed by non-calcified, mixed, and calcified plaque, respectively (log-rank p < 0.001). Among the prediction models for MACE, adding stenosis degree to FRS showed better discrimination and risk reclassification compared to FRS or the FRS + CACS model (all p < 0.05). Furthermore, incorporating plaque type in the prediction model significantly improved reclassification (integrated discrimination improvement, 0.08; p = 0.023) and showed the highest discrimination index (C-statistics, 0.85). However, the addition of CACS on CCTA with FRS did not add to the prediction ability for MACE (p > 0.05). Conclusion: Assessment of stenosis degree and plaque type using CCTA provided additional prognostic value over CACS and FRS to risk stratify stroke patients without prior history of CAD better.

• Journal of Korea Multimedia Society
• /
• v.23 no.9
• /
• pp.1155-1163
• /
• 2020

Deep learning approach is emerging as a new way to improve the accuracy of tree species identification using bark image. However, the approach has not been studied enough because it is confronted with the problem of acquiring a large volume of bark image dataset. This study solved this problem by utilizing a pretrained off-the-shelf DCNN model. It compares the discrimination power of bark features extracted by each DCNN model. Then it extracts the features by using a selected DCNN model and feeds them to a multi-layer perceptron (MLP). We found out that the ResNet50 model is effective in extracting bark features and the MLP could be trained well with the features reduced by the principal component analysis. The proposed approach gives accuracy of 99.1% and 98.4% for BarkTex and Trunk12 datasets respectively.

• Journal of Information Technology Applications and Management
• /
• v.11 no.1
• /
• pp.161-174
• /
• 2004

This study is intended to suggest1 the optimized data mining model for the efficient customer credit evaluation in the capital finance industry. To accomplish the research objective, various data mining models for the customer credit evaluation are compared and analyzed. Furthermore, existing models such as Multi-Layered Perceptrons, Multivariate Discrimination Analysis, Radial Basis Function, Decision Tree, and Logistic Regression are employed for analyzing the customer information in the capital finance market and the detailed data of capital financing transactions. Finally, the data from the integrated model utilizing a genetic algorithm is compared with those of each individual model mentioned above. The results reveals that the integrated model is superior to other existing models.

• Journal of international Conference on Electrical Machines and Systems
• /
• v.3 no.3
• /
• pp.325-330
• /
• 2014

Eddy current induction is used in a wide range of electronic devices, for example in detection sensors. Due to the advances in computer hardware and software, the need for 3D computation and system comprehension is a requirement to develop and optimize such devices nowadays. Pure theoretical models are mostly limited to special cases. On the other hand, the classical use of commercial Finite Element (FE) electromagnetic 3D models is not computationally efficient and lacks modeling flexibility or robustness. The proposed approach focuses on: (1) implementing theoretical formulations in 3D (FE) model of a detection device as well as (2) an automatic Volumetric Estimation Method (VEM) developed to selectively model the target finite elements. Due to these two approaches, this model is suitable for parametric studies and optimization of the number, location, shape, and size of PCB receivers in order to get the desired target discrimination information preserving high accuracy with tenfold reduction in computation time compared to commercial FE software.

• Journal of Biomedical Engineering Research
• /
• v.9 no.1
• /
• pp.1-10
• /
• 1988

In this study, the recognition experiments for Korean vowel are performed using peripheral auditory model. In addition, for the purpose of objective comparison, the recognition experiments are performed by extracting LPC cepstrum coefficients for the same speech data. The results are as follows. 1) The time and the frequency responses of the auditory model show that important features of input signal are involved in the responses of inner ear and auditory nerve. 2) The recognition results for Korean vowel show that the recognition rate by auditory model output is higher than the recognition rate by LPC cepstrum coefficients. 3) The adaptation phenomenon of auditory nerve provides useful characteristics for the discrimination of vowel signal.

• Communications for Statistical Applications and Methods
• /
• v.21 no.6
• /
• pp.501-512
• /
• 2014

Stylized facts on asset return are fat-tail, asymmetry, volatility clustering and structure changes. This paper simultaneously captures these characteristics by introducing a multi-regime models: Finite mixture distribution and regime switching GARCH model. Analyzing the daily KOSPI return from $4^{th}$ January 2000 to $30^{th}$ June 2014, we find that a two-component mixture of t distribution is a good candidate to describe the shape of the KOSPI return from unconditional and conditional perspectives. Empirical results suggest that the equality assumption on the shape parameter of t distribution yields better discrimination of heterogeneity component in return data. We report the strong regime-dependent characteristics in volatility dynamics with high persistence and asymmetry by employing a regime switching GJR-GARCH model with t innovation model. Compared to two sub-samples, Pre-Crisis (January 2003 ~ December 2007) and Post-Crisis (January 2010 ~ June 2014), we find that the degree of persistence in the Pre-Crisis is higher than in the Post-Crisis along with a strong asymmetry in the low-volatility (high-volatility) regime during the Pre-Crisis (Post-Crisis).

• Smart Media Journal
• /
• v.5 no.3
• /
• pp.30-34
• /
• 2016

Nowadays, there are a lot of Korean documents, which often need to be identified in one of printed or handwritten text. Early methods for the identification use structural features, which can be simple and easy to apply to text of a specific font, but its performance depends on the font type and characteristics of the text. Recently, the bag-of-words model has been used for the identification, which can be invariant to changes in font size, distortions or modifications to the text. The method based on bag-of-words model includes three steps: word segmentation using connected component grouping, feature extraction, and finally classification using SVM(Support Vector Machine). In this paper, bag-of-words model based method is proposed using SURF(Speeded Up Robust Feature) for the identification of machine printed and handwritten text in Korean documents. The experiment shows that the proposed method outperforms methods based on structural features.