• Journal of KIISE:Software and Applications
• /
• v.32 no.8
• /
• pp.781-788
• /
• 2005

LDA (Linear Discriminant Analysis) is a data discrimination technique that seeks transformation to maximize the ratio of the between-class scatter and the within-class scatter While it has been successfully applied to several applications, it has two limitations, both concerning the underfitting problem. First, it fails to discriminate data with complex distributions since all data in each class are assumed to be distributed in the Gaussian manner; and second, it can lose class-wise information, since it produces only one transformation over the entire range of classes. We propose three extensions of LDA to overcome the above problems. The first extension overcomes the first problem by modeling the within-class scatter using a PCA mixture model that can represent more complex distribution. The second extension overcomes the second problem by taking different transformation for each class in order to provide class-wise features. The third extension combines these two modifications by representing each class in terms of the PCA mixture model and taking different transformation for each mixture component. It is shown that all our proposed extensions of LDA outperform LDA concerning classification errors for handwritten digit recognition and alphabet recognition.

• The Transactions of the Korea Information Processing Society
• /
• v.6 no.1
• /
• pp.253-260
• /
• 1999

In this paper, we suggest a new direct method for mage segmentation using texture and color information combined through a multivariate linear discriminant algorithm. The color texture is computed in nin 3${\times}$3 masks obtained from each 3${\times}$3${\times}$3 spatio-spectral neighborhood in the image using the classical haralick and Pressman texture features. Among these 9${\times}$28 texture features the best set was extracted from a training set. The resulting set of 10 features were used to segment an image into four different regions. The resulting segmentation was Compared to classical color and texture segmentation methods using both box classifiers and maximum likelihood classification. It compared favourably on the test image from a Fastred-Lightgreen stained prostatic histological tissue section based on visual inspection. The classification accuracy of 97.5% for the new method obtained on the training data was also among the best of the tested methods. If these results hold for a larger set of images, this method should be a useful tool for segmenting images where both color and texture are relevant for the segmentation process.

• The KIPS Transactions:PartD
• /
• v.19D no.2
• /
• pp.147-150
• /
• 2012

While LDA is a supervised dimension reduction method which finds projective directions to maximize separability between classes, the performance of LDA is severely degraded when the number of labeled data is small. Recently semi-supervised dimension reduction methods have been proposed which utilize abundant unlabeled data and overcome the shortage of labeled data. However, matrix computation usually used in statistical dimension reduction methods becomes hindrance to make the utilization of a large number of unlabeled data difficult, and moreover too much information from unlabeled data may not so helpful compared to the increase of its processing time. In order to solve these problems, we propose an ensemble approach for semi-supervised dimension reduction. Extensive experimental results in text classification demonstrates the effectiveness of the proposed method.

• Journal of the Korea Society of Computer and Information
• /
• v.19 no.7
• /
• pp.1-8
• /
• 2014

Clustering is one of the most popular unsupervised learning methods, which is widely used to form clusters with homogeneous data. Clustering was used to extract contexts corresponding to clusters and a classification method was applied to each context or cluster individually. However, it is difficult to say that the unsupervised clustering is the best context forming method from the view of classification. In this paper, a new clustering method considering classification was proposed. The proposed method tries to minimize classification error in each cluster when a classification method is applied to each context locally. For this purpose, the proposed method adds constraints forcing two data points belong to the same class to have small distances, and two data points belong to different classes to have large distances in each cluster like in linear discriminant analysis. The usefulness of the proposed method is confirmed by experimental results.

• Journal of the Korean housing association
• /
• v.25 no.1
• /
• pp.51-61
• /
• 2014

The purpose of this study was to explore housing characteristics of young single- or two-person households in the U.S. metropolitan urban areas and determinants of their housing cost burden. Total 764 single-person households, 744 two-person households and 424 households with three or more persons were selected from the 2011 American Housing Survey public-use microdata for the study based on specific sampling criteria. The major findings are as follows: (1) In comparisons with larger households, single- or two-person households were characterized to be headed by younger householders, to have less income, and to have a greater proportion of households living in central cities of metropolitan areas, renting housing units, living in smaller size units or multifamily structures; (3) housing cost of single- or two-person households were significantly less than a larger households while housing costs per unit square footage (SQFT) of single- or two-person households was significantly greater; (4) regardless of the household size, there are many household headed by young college graduates paying too much of their income for housing, and single-person households were found to have the greatest housing cost burden; and (5) a linear combination of low-income status, monthly housing costs per unit SQFT, annual household income, and unit SQFT per person was found to be most efficient to predict single- or two-person households with housing cost burden.

• Journal of Institute of Control, Robotics and Systems
• /
• v.22 no.3
• /
• pp.199-203
• /
• 2016

A brain-computer interface or BCI provides an alternative method for acting on the world. Brain signals can be recorded from the electrical activity along the scalp using an electrode cap. By analyzing the EEG, it is possible to determine whether a person is thinking about his/her hand or foot movement and this information can be transferred to a machine and then translated into commands. However, we do not know which information relates to motor imagery and which channel is good for extracting features. A general approach is to use all electronic channels to analyze the EEG signals, but this causes many problems, such as overfitting and problems removing noisy and artificial signals. To overcome these problems, in this paper we used a new optimization method called the Fruit Fly optimization algorithm (FOA) to select the best channels and then combine them with CSP method to extract features to improve the classification accuracy by linear discriminant analysis. We also used particle swarm optimization (PSO) and a genetic algorithm (GA) to select the optimal EEG channel and compared the performance with that of the FOA algorithm. The results show that for some subjects, the FOA algorithm is a better method for selecting the optimal EEG channel in a short time.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.11 no.12
• /
• pp.6069-6091
• /
• 2017

Augmented Reality (AR) overlays virtual information on real world data, such as displaying useful information on videos/images of a scene. This paper presents an Enhanced AR (EAR) system that displays useful statistical players' information on captured images of a sports game. We focus on the situation where the input image is degraded by strong sunlight. Proposed EAR system consists of an image enhancement technique to improve the accuracy of subsequent player and face detection. The image enhancement is followed by player and face detection, face recognition, and players' statistics display. First, an algorithm based on multi-scale retinex is proposed for image enhancement. Then, to detect players' and faces', we use adaptive boosting and Haar features for feature extraction and classification. The player face recognition algorithm uses boosted linear discriminant analysis to select features and nearest neighbor classifier for classification. The system can be adjusted to work in different types of sports where the input is an image and the desired output is display of information nearby the recognized players. Simulations are carried out on 2096 different images that contain players in diverse conditions. Proposed EAR system demonstrates the great potential of computer vision based approaches to develop AR applications.

• Journal of Biomedical Engineering Research
• /
• v.35 no.4
• /
• pp.75-80
• /
• 2014

This paper proposed a method that can automatically classify sleep apnea by using features extracted from pulse rate variability(PRV) signals induced from piezo snoring sensor for patients with obstructive sleep apnea(OSA). We have extracted eight features(NN, SDNN, RMSSD, NN10, NN50, LF, HF and LF/HF ratio) based on time and frequency analyses of PRV. Sleep apnea was classified by a linear discriminant analysis(LDA). A performance was evaluated using snore recordings from 13 patients with OSA (ages: $54.5{\pm}10.5$ years, body mass index: $26.3{\pm}2.5kg/m^2$, apnea-hypopnea index: $19.2{\pm}6.0/h$). The sensitivity and specificity were $78.9{\pm}0.9%$ and $78.9{\pm}0.9%$ for training set and $77.7{\pm}10.9%$ and $79.0{\pm}2.8%$ for test set, respectively. Our study demonstrated the feasibility of implementing a piezo snoring sensor based on a portable device as a simple and cost-effective solution for contributing to the OSA screening.

• Journal of the Institute of Electronics Engineers of Korea CI
• /
• v.45 no.5
• /
• pp.15-24
• /
• 2008

For high-dimensional classification tasks, such as face recognition, the number of samples is smaller than the dimensionality of the samples. In such cases, a problem encountered in linear discriminant analysis-based methods for dimension reduction is what is known as the small sample size (SSS) problem. Recently, to solve the SSS problem, a way of employing a dissimilarity-based classification(DBC) has been investigated. In DBC, an object is represented based on the dissimilarity measures among representatives extracted from training samples instead of the feature vector itself. In this paper, we propose a new method of optimizing DBCs using multi-level fusion strategies(MFS), in which fusion strategies are employed to represent features as well as to design classifiers. Our experimental results for benchmark face databases demonstrate that the proposed scheme achieves further improved classification accuracies.

• Journal of Biomedical Engineering Research
• /
• v.40 no.6
• /
• pp.242-249
• /
• 2019

Forearm electromyography (EMG) generated by wrist movements has been widely used to develop an electrical prosthetic hand, but EMG generated by finger movements has been rarely used even though 20% of amputees lose fingers. The goal of this study is to improve the classification performance of different finger movements using a deep learning algorithm, and thereby contributing to the development of a high-performance finger-based prosthetic hand. Ten participants took part in this study, and they performed seven different finger movements forty times each (thumb, index, middle, ring, little, fist and rest) during which EMG was measured from the back of the right hand using four bipolar electrodes. We extracted mean absolute value (MAV), root mean square (RMS), and mean (MEAN) from the measured EMGs for each trial as features, and a 5x5-fold cross-validation was performed to estimate the classification performance of seven different finger movements. A long short-term memory (LSTM) model was used as a classifier, and linear discriminant analysis (LDA) that is a widely used classifier in previous studies was also used for comparison. The best performance of the LSTM model (sensitivity: 91.46 ± 6.72%; specificity: 91.27 ± 4.18%; accuracy: 91.26 ± 4.09%) significantly outperformed that of LDA (sensitivity: 84.55 ± 9.61%; specificity: 84.02 ± 6.00%; accuracy: 84.00 ± 5.87%). Our result demonstrates the feasibility of a deep learning algorithm (LSTM) to improve the performance of classifying different finger movements using EMG.