• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.13 no.4
• /
• pp.2162-2179
• /
• 2019

The performance of text classification is highly related to the feature selection methods. Usually, two tasks are performed when a feature selection method is applied to construct a feature set; 1) assign score to each feature and 2) select the top-N features. The selection of top-N features in the existing filter-based feature selection methods is biased by their discriminative power and the empirical process which is followed to determine the value of N. In order to improve the text classification performance by presenting a more illustrative feature set, we present an approach via a potent representation learning technique, namely DBN (Deep Belief Network). This algorithm learns via the semantic illustration of documents and uses feature vectors for their formulation. The nodes, iteration, and a number of hidden layers are the main parameters of DBN, which can tune to improve the classifier's performance. The results of experiments indicate the effectiveness of the proposed method to increase the classification performance and aid developers to make effective decisions in certain domains.

• Journal of the Institute of Electronics and Information Engineers
• /
• v.52 no.10
• /
• pp.129-136
• /
• 2015

We propose a hybrid face recognition to deal with illumination variation. For this, we extract discriminant features by using the different illumination invariant feature extraction methods. In order to utilize both advantages of each method, we evaluate the discriminant power of each feature by using the discriminant distance and then construct a composite feature with only the features that contain a large amount of discriminative information. The experimental results for the Multi-PIE, Yale B, AR and yale databases show that the proposed method outperforms an individual illumination invariant feature extraction method for all the databases.

• Journal of Electrical Engineering and Technology
• /
• v.7 no.2
• /
• pp.245-254
• /
• 2012

In this study, we introduce a prototype-based classifier with feature selection that dwells upon the usage of a biologically inspired optimization technique of Particle Swarm Optimization (PSO). The design comprises two main phases. In the first phase, PSO selects P % of patterns to be treated as prototypes of c classes. During the second phase, the PSO is instrumental in the formation of a core set of features that constitute a collection of the most meaningful and highly discriminative coordinates of the original feature space. The proposed scheme of feature selection is developed in the wrapper mode with the performance evaluated with the aid of the nearest prototype classifier. The study offers a complete algorithmic framework and demonstrates the effectiveness (quality of solution) and efficiency (computing cost) of the approach when applied to a collection of selected data sets. We also include a comparative study which involves the usage of genetic algorithms (GAs). Numerical experiments show that a suitable selection of prototypes and a substantial reduction of the feature space could be accomplished and the classifier formed in this manner becomes characterized by low classification error. In addition, the advantage of the PSO is quantified in detail by running a number of experiments using Machine Learning datasets.

• ETRI Journal
• /
• v.41 no.2
• /
• pp.235-241
• /
• 2019

In this paper, we propose a rank-weighted reconstruction feature to improve the robustness of a feed-forward deep neural network (FFDNN)-based acoustic model. In the FFDNN-based acoustic model, an input feature is constructed by vectorizing a submatrix that is created by slicing the feature vectors of frames within a context window. In this type of feature construction, the appropriate context window size is important because it determines the amount of trivial or discriminative information, such as redundancy, or temporal context of the input features. However, we ascertained whether a single parameter is sufficiently able to control the quantity of information. Therefore, we investigated the input feature construction from the perspectives of rank and nullity, and proposed a rank-weighted reconstruction feature herein, that allows for the retention of speech information components and the reduction in trivial components. The proposed method was evaluated in the TIMIT phone recognition and Wall Street Journal (WSJ) domains. The proposed method reduced the phone error rate of the TIMIT domain from 18.4% to 18.0%, and the word error rate of the WSJ domain from 4.70% to 4.43%.

• International Journal of Contents
• /
• v.8 no.4
• /
• pp.56-63
• /
• 2012

In recent decades, analyzing the activities of human brain achieved some accomplishments by using the functional Magnetic Resonance Imaging (fMRI) technique. fMRI data provide a sequence of three-dimensional images related to human brain's activity which can be used to detect instantaneous cognitive states by applying machine learning methods. In this paper, we propose a new approach for distinguishing human's cognitive states such as "observing a picture" versus "reading a sentence" and "reading an affirmative sentence" versus "reading a negative sentence". Since fMRI data are high dimensional (about 100,000 features in each sample), extremely sparse and noisy, feature selection is a very important step for increasing classification accuracy and reducing processing time. We used the Fisher Discriminant Ratio to select the most powerful discriminative features from some Regions of Interest (ROIs). The experimental results showed that our approach achieved the best performance compared to other feature extraction methods with the average accuracy approximately 95.83% for the first study and 99.5% for the second study.

• Journal of the Korea Society of Computer and Information
• /
• v.20 no.11
• /
• pp.9-16
• /
• 2015

We propose a hybrid pattern recognition method that effectively combines two different features for improving data classification. We first extract the PCA (Principal Component Analysis) and LDA (Linear Discriminant Analysis) features, both of which are widely used in pattern recognition, to construct a set of basic features, and then evaluate the separability of each basic feature. According to the results of evaluation, we select only the basic features that contain a large amount of discriminative information for construction of the combined features. The experimental results for the various data sets in the UCI machine learning repository show that using the proposed combined features give better recognition rates than when solely using the PCA or LDA features.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.11 no.3
• /
• pp.1670-1683
• /
• 2017

Mitotic event recognition is a crucial and challenging task in biomedical applications. In this paper, we introduce the slow feature analysis and propose a fully-automated mitotic event recognition method for cell populations imaged with time-lapse phase contrast microscopy. The method includes three steps. First, a candidate sequence extraction method is utilized to exclude most of the sequences not containing mitosis. Next, slow feature is learned from the candidate sequences using slow feature analysis. Finally, a hidden conditional random field (HCRF) model is applied for the classification of the sequences. We use a supervised SFA learning strategy to learn the slow feature function because the strategy brings image content and discriminative information together to get a better encoding. Besides, the HCRF model is more suitable to describe the temporal structure of image sequences than nonsequential SVM approaches. In our experiment, the proposed recognition method achieved 0.93 area under curve (AUC) and 91% accuracy on a very challenging phase contrast microscopy dataset named C2C12.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.18 no.6
• /
• pp.1540-1561
• /
• 2024

This research proposes a novel approach for Tamil Handwritten Character Recognition (THCR) that combines feature selection and ensemble learning techniques. The Tamil script is complex and highly variable, requiring a robust and accurate recognition system. Feature selection is used to reduce dimensionality while preserving discriminative features, improving classification performance and reducing computational complexity. Several feature selection methods are compared, and individual classifiers (support vector machines, neural networks, and decision trees) are evaluated through extensive experiments. Ensemble learning techniques such as bagging, and boosting are employed to leverage the strengths of multiple classifiers and enhance recognition accuracy. The proposed approach is evaluated on the HP Labs Dataset, achieving an impressive 95.56% accuracy using an ensemble learning framework based on support vector machines. The dataset consists of 82,928 samples with 247 distinct classes, contributed by 500 participants from Tamil Nadu. It includes 40,000 characters with 500 user variations. The results surpass or rival existing methods, demonstrating the effectiveness of the approach. The research also offers insights for developing advanced recognition systems for other complex scripts. Future investigations could explore the integration of deep learning techniques and the extension of the proposed approach to other Indic scripts and languages, advancing the field of handwritten character recognition.

• Proceedings of the KSPS conference
• /
• 2002.11a
• /
• pp.115-118
• /
• 2002

Neural networks have been known to have great discriminative power in pattern classification problems. In this paper, the multilayer perceptron neural networks are employed to automatically detect laryngeal pathology in speech. Also new feature parameters are introduced which can reflect the periodicity of speech and its perturbation. These parameters and cepstral coefficients are used as input of the multilayer perceptron neural networks. According to the experiment using Korean disordered speech database, incorporation of new parameters with cepstral coefficients outperforms the case with only cepstral coefficients.

• Proceedings of the KIEE Conference
• /
• 2015.07a
• /
• pp.1391-1392
• /
• 2015

최근 컴퓨터 기술이 발전하고, 컴퓨터의 형태가 다양해지면서 여러 wearable device들이 생겨났다. 이에 따라 휴먼 인터페이스 기술에서 사람의 감정정보가 중요해졌고, 감정인식에 대한 연구들이 많이 진행 되어 왔다. 본 논문에서는 감정분석에 적합한 특징벡터를 제시하고자 한다. 이를 위해 사람의 감정을 보통, 기쁨, 슬픔, 화남 4가지로 분류하고 방송매체를 통하여 잡음 없이 녹음하였다. 특징벡터는 MFCC, LPC, LPCC 3가지를 추출하였고 Bhattacharyya거리 측정을 통하여 분리도를 비교하였다.