• Proceedings of the Acoustical Society of Korea Conference
• /
• 1994.06c
• /
• pp.138-141
• /
• 1994

기존의 HMM을 이용한 음성인식기는 대부분 ML 추정에 기초한 Baum-Welch 알고리듬으로 학습되었다. ML학습은 기본적으로 무한한 양의 학습 데이터가 주어지고, 각 모델들이 서로 독립이라는 가정에 기초한다. 하지만 실제적인 학습의 경우에 각 모델들이 서로 독립이라고 보기 어렵고, 학습 데이터의 양도 상당히 제한되어 있어서 인식기의 변별력을 저하시키는 주된 원인이 되고 있다. 본 논문에서는 전통적인 패턴분류기법인 Bayes 결정이론에 따라 최소오차율분류를 위한 MAP 수식화를 유도하고, 그에 기초한 HMM의 변별력 있는 학습 알고리듬을 제안한다. 최소오차율분류를 근사화한 사후확률로 표현된 비용함수를 정의하고, 그 비용함수에 조건부 경사강하법을 적용한다. 제안된 알고리듬을 분류하기 어려운 한국어 단음절 인식에 적용한 결과, 기존의 ML 알고리듬으로 학습한 경우 발생한 오인식 개수의 약 10% 가량이 개선되었다.

• Annual Conference on Human and Language Technology
• /
• 1990.11a
• /
• pp.172-179
• /
• 1990

본 논문에서는 한글(한국어) ascii 코드와 4가지 한글 표현 원리에 만들어지는 한글음절을 정의한다. Common Lisp(CL)의 확장된 버전으로 한글이 사용 가능한 한글 CL(HCL)의 소개하고 CL에 추가되는 새로운 한글 함수에 대하여 설명한다. HCL의 모든 함수는 한글을 다루는 방법에 따라 4가지 타입으로 나뉘어진다. 1) 타입 0 - 한글을 입출력 값으로 취하지 않는 전형적인 CL 함수, 2) 타입 1 - 원래 CL 함수정의의 변경없이 입력으로 한글을 받아들이는 함수, 3) 타입 2 - 한글을 사용하기 위하여 함수의 정의를 확장해야하는 CL 함수, 4) 타입 3 - 한글 처리를 하기 위하여 새로 설계한 새로운 함수. 위의 타입에 의해 분류되는 각 함수에 대한 정의를 제안하고 한글 편집기에 대하여 소개한다.

• The Korean Journal of Applied Statistics
• /
• v.33 no.4
• /
• pp.395-407
• /
• 2020

Most classification accuracy measures for optimal threshold are divided into two types: one is expressed with cumulative distribution functions and probability density functions, the other is based on ROC curve and AUC. Unal (2017) proposed the index of union (IU) as an accuracy measure that considers two types to get them. In this study, ten kinds of accuracy measures (including IU) are divided into six categories, and the advantages of the IU are studied by comparing the measures belonging to each category. The optimal thresholds of these measures are obtained by setting various normal mixture distributions; subsequently, the first and second type of errors as well as the error sums corresponding to each threshold are calculated. The properties and characteristics of the IU statistic are explored by comparing the discriminative power of other accuracy measures based on error values.The values of the first type error and error sum of IU statistic converge to those of the best accuracy measures of the second category as the mean difference between the two distributions increases. Therefore, IU could be an accuracy measure to evaluate the discriminant power of a model.

• Journal of KIISE
• /
• v.41 no.12
• /
• pp.1050-1057
• /
• 2014

Electroencephalogram (EEG)-based brain-computer interfaces (BCI) can be used for a number of purposes in a variety of industries, such as to replace body parts like hands and feet or to improve user convenience. In this paper, we propose a method to decompose and extract motor imagery EEG signal using Empirical Mode Decomposition (EMD) and Fast Fourier Transforms (FFT). The EEG signal classification consists of the following three steps. First, during signal decomposition, the EMD is used to generate Intrinsic Mode Functions (IMFs) from the EEG signal. Then during feature extraction, the power spectral density (PSD) is used to identify the frequency band of the IMFs generated. The FFT is used to extract the features for motor imagery from an IMF that includes mu rhythm. Finally, during classification, the Support Vector Machine (SVM) is used to classify the features of the motor imagery EEG signal. 10-fold cross-validation was then used to estimate the generalization capability of the given classifier., and the results show that the proposed method has an accuracy of 84.50% which is higher than that of other methods.

• Journal of KIISE:Software and Applications
• /
• v.32 no.1
• /
• pp.1-9
• /
• 2005

In this paper we propose a modified fuzzy min-max neural network model for pattern classification and discuss the usefulness of the model. We define a new hypercube membership function which has a weight factor to each of the feature within a hyperbox. The weight factor makes it possible to consider the degree of relevance of each feature to a class during the classification process. Based on the proposed model, a knowledge extraction method is presented. In this method, a list of relevant features for a given class is extracted from the trained network using the hyperbox membership functions and connection weights. Ft)r this purpose we define a Relevance Factor that represents a degree of relevance of a feature to the given class and a similarity measure between fuzzy membership functions of the hyperboxes. Experimental results for the proposed methods and discussions are presented for the evaluation of the effectiveness and feasibility of the proposed methods.

• Journal of the Korean Geotechnical Society
• /
• v.33 no.1
• /
• pp.57-65
• /
• 2017

During a construction of embankment, sub base, or retaining wall backfill, the speedy measurement of water content is necessary. In this study, a test method for field determination of water content of soil by the calcium carbide gas pressure (speedy water content measurement method) was evaluated for its reliability and accuracy. Dry oven and microwave oven methods were also used for water content measurement. In the test, weathered granite and Nakdong River sand in the site and kaolinite were used for water content measurement. The mass of 20, 22, 24, 26, 28, and 30 g of soil was respectively tested for 1, 3, and 5 min. The effect of each sample on water content was compared one another and analyzed. As the mass and testing time increased, the water content increased. The amount of soil was more important factor than testing time for the speedy water content measurement. In order to obtain similar result to that of dry oven method, 3 min of testing time with 24 g of soil was necessary for weathered granite classified as SM and 3 min with 30 g for Nakdong River sand classified as SP. For Nakdong River sand with 20-50% of kaolinite, the water content by speedy measurement increased as the clay content increased.

• The Journal of the Acoustical Society of Korea
• /
• v.23 no.4
• /
• pp.296-302
• /
• 2004

Acoustic experiments are performed for a seafloor classification from 19 May to 25 May 2003. The six different sites of bottom composition are settled and the bottom reflection losses with frequencies (30, 50, 80. 100, 120 kHz) are measured. Sediment samples were collected using gravity core and the sample was extracted for grain size analysis. The fuzzy logic is used to classify the seabed. In the fuzzy logic. Bottom 1083 model of frequency dependence is used as the input membership functions and the output membership functions are composed of the Wentworth grain size of the bottom. The possibility of the seafloor classification is verified comparing the inversed mean grain size using fuzzy logic with the results of the coring.

• The Korean Journal of Applied Statistics
• /
• v.35 no.1
• /
• pp.77-91
• /
• 2022

The optimal threshold estimation is considered in order to discriminate the mixture distribution in the fields of Biostatistics and credit evaluation. There exists well-known various accuracy measures that examine the discriminant power. Recently, Matthews correlation coefficient and the F1 statistic were studied to estimate optimal thresholds. In this study, we explore whether these accuracy measures are appropriate for the optimal threshold to discriminate the mixture distribution. It is found that some accuracy measures that depend on the sample size are not appropriate when two sample sizes are much different. Moreover, an alternative method for finding the optimal threshold is proposed using the correlation coefficient that defines the ratio of the confusion matrix, and the usefulness and utility of this method are also discusses.

• The Korean Journal of Applied Statistics
• /
• v.21 no.3
• /
• pp.547-560
• /
• 2008

It is well known that spectral density function determines auto-covariance function of stationary time-series data and smoothed periodogram is a consistent estimator of spectral density function. Recently, Kim and Park (2007) showed that smoothed- periodogram based distances performs very well for the classification. In this paper, we introduce classification methods with smoothed periodogram and apply the approaches to the monthly precipitation measurements obtained from January, 1987 through December, 2007 at 22 locations in South Korea.

• Asia-Pacific Journal of Business Venturing and Entrepreneurship
• /
• v.11 no.4
• /
• pp.247-264
• /
• 2016

The purpose of this study is to find the determinants for the type(Franchising or Independent Entrepreneurship) of new venture creation. This study conducted an empirical analysis on a total of 398 samples of survey gathered from people in Seoul, Gyoeng-gi, Daegu, and Gyeonsangbuk-do. This study includes not only personal traits, but also entrepreneurial intention and network as independent variables. Findings of the analysis reported that Entrepreneurial intention, Need for achievement, Autonomy, Entrepreneurship, Self-efficacy, Education, Network, Age, and Income have significant discriminant power, in order of priority, on general two groups of Franchising and Independent Entrepreneurship. However, in the study, autonomy is revealed as the sole discriminant factor on considering venture creation groups. Based on the result, the study contributes theoretical and practical implications in relation to the determinants for the type of franchising or independent entrepreneurship.