• Proceedings of the KSRS Conference
• /
• 2007.10a
• /
• pp.6-10
• /
• 2007

Classification of hyperspectral images is challenging. A very high dimensional input space requires an exponentially large amount of data to adequately and reliably represent the classes in that space. In other words in order to obtain statistically reliable classification results, the number of necessary training samples increases exponentially as the number of spectral bands increases. However, in many situations, acquisition of the large number of training samples for these high-dimensional datasets may not be so easy. This problem can be overcome by using multiple classifiers. In this paper we compared the effectiveness of two approaches for creating multiple classifiers, feature selection and feature extraction. The methods are based on generating multiple feature subsets by running feature selection or feature extraction algorithm several times, each time for discrimination of one of the classes from the rest. A maximum likelihood classifier is applied on each of the obtained feature subsets and finally a combination scheme was used to combine the outputs of individual classifiers. Experimental results show the effectiveness of feature extraction algorithm for generating multiple classifiers.

• Journal of Information Processing Systems
• /
• v.14 no.4
• /
• pp.807-823
• /
• 2018

Speaker verification system performance depends on the utterance of each speaker. To verify the speaker, important information has to be captured from the utterance. Nowadays under the constraints of limited data, speaker verification has become a challenging task. The testing and training data are in terms of few seconds in limited data. The feature vectors extracted from single frame size and rate (SFSR) analysis is not sufficient for training and testing speakers in speaker verification. This leads to poor speaker modeling during training and may not provide good decision during testing. The problem is to be resolved by increasing feature vectors of training and testing data to the same duration. For that we are using multiple frame size (MFS), multiple frame rate (MFR), and multiple frame size and rate (MFSR) analysis techniques for speaker verification under limited data condition. These analysis techniques relatively extract more feature vector during training and testing and develop improved modeling and testing for limited data. To demonstrate this we have used mel-frequency cepstral coefficients (MFCC) and linear prediction cepstral coefficients (LPCC) as feature. Gaussian mixture model (GMM) and GMM-universal background model (GMM-UBM) are used for modeling the speaker. The database used is NIST-2003. The experimental results indicate that, improved performance of MFS, MFR, and MFSR analysis radically better compared with SFSR analysis. The experimental results show that LPCC based MFSR analysis perform better compared to other analysis techniques and feature extraction techniques.

• KIPS Transactions on Software and Data Engineering
• /
• v.2 no.1
• /
• pp.43-48
• /
• 2013

Adaboost is widely used for Haar-like feature boosting algorithm in Face Detection. It shows very effective performance on single distribution model. But when detecting front and side face images at same time, Adaboost shows it's limitation on multiple distribution data because it uses linear combination of basic classifier. This paper suggest the HDCT, modified decision tree algorithm for Haar-like features. We still tested the performance of HDCT compared with Adaboost on multiple distributed image recognition.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.18 no.6
• /
• pp.871-875
• /
• 2008

This paper deals with the feature selection for multiple k-nearest neighbor (k-NN) classifiers using Genetic Algorithm with Varying reputation Size (GAVaPS). Because we use multiple k-NN classifiers, the feature selection problem for them is vary hard and has large search region. To solve this problem, we employ the GAVaPS which outperforms comparison with simple genetic algorithm (SGA). Further, we propose the efficient combining method for multiple k-NN classifiers using GAVaPS. Experiments are performed to demonstrate the efficiency of the proposed method.

• Journal of KIISE:Software and Applications
• /
• v.30 no.7_8
• /
• pp.659-671
• /
• 2003

This paper proposes a method of disambiguating the senses of words using decision lists, which consists of rules with confidence values. The rule of decision list is composed of a boolean function(=precondition) and a class(=sense). Decision lists classify the instance using the rule with the highest confidence value that is matched with it. Previous work disambiguated the senses using single feature decision lists, whose boolean function was composed of only one feature. However, this approach can be affected more severely by data sparseness problem and preprocessing errors. Hence, we propose multiple feature decision lists that have the boolean function consisting of more than one feature in order to identify the senses of words. Experiments are performed with 1 sense tagged corpus in Korean and 5 sense tagged corpus in English. The experimental results show that multiple feature decision lists are more effective than single feature decision lists in disambiguating senses.

• Journal of Korea Multimedia Society
• /
• v.19 no.5
• /
• pp.890-900
• /
• 2016

This paper presents a multiple target tracking system using target feature information. In the proposed system, the state of target is defined as its kinematic as well as feature : the kinematic includes a location and a velocity; the feature contains the image correlation between a prior target and a current measurement. The feature information is used for generating the validation matrix and association probability of joint probabilistic data association (JPDA) algorithm. Through the Kalman filter, the target kinematic is updated. Then the tracking information is cycled by the track management algorithm. The system has been evaluated using the images obtained from Electro-Optics/ InfraRed (EO/IR) sensor. It is verified that the proposed system can reduce the complexity burden of JPDA process and can enhance the track maintenance rate.

• Journal of the Korea Society of Computer and Information
• /
• v.19 no.1
• /
• pp.23-30
• /
• 2014

Feature matching using image descriptors is robust method used recently. However, mismatches occur in 3D transformed images, illumination-changed images and repetitive-pattern images. In this paper, we observe that there are a lot of mismatches in the images which have repetitive patterns. We analyze it and propose a method to eliminate these mismatches. MDMF(Multiple Descriptors-based Mismatch Filtering) eliminates mismatches by using descriptors of nearest several features of one specific feature point. In experiments, for geometrical transformation like scale, rotation, affine, we compare the match ratio among SIFT, ASIFT and MDMF, and we show that MDMF can eliminate mismatches successfully.

• The Journal of Korean Institute of Electromagnetic Engineering and Science
• /
• v.27 no.5
• /
• pp.407-415
• /
• 2016

Recent wireless communication trends have emphasized the importance of energy-efficient transmission. In this paper, link adaptation with machine learning mechanism for maximum energy efficiency in multiple-input multiple-output orthogonal frequency division multiplexing(MIMO-OFDM) wireless system is considered. For reflecting frequency-selective MIMO-OFDM channels, two-dimensional capacity(2D-CAP) feature space is proposed. In addition, machine-learning-based bit and power adaptation(ML-BPA) algorithm that performs classification-based link adaptation is presented. Simulation results show that 2D-CAP feature space can represent channel conditions accurately and bring noticeable improvement in link adaptation performance. Compared with other feature spaces, including ordered postprocessing signal-to-noise ratio(ordSNR) feature space, 2D-CAP has distinguished advantages in either efficiency performance or computational complexity.

• Journal of information and communication convergence engineering
• /
• v.9 no.4
• /
• pp.466-470
• /
• 2011

This paper proposes the new method based on fuzzy logic which recognizes between normal and abnormal. The object image was the Thyroid Gland cell image that was diagnosed as normal and abnormal(two types of abnormal : follicular neoplastic cell, and papillary neoplastic cell), respectively. The nuclei were successfully diagnosed as normal and abnormal. The multiple feature parameters (pre-obtained 16 feature parameters of image data) were used to extract the features of each nucleus. As a consequence of using fuzzy logic algorithm, proposed in this paper, average recognition rate of 98.25% was obtained.

• Proceedings of the KSRS Conference
• /
• v.2
• /
• pp.772-775
• /
• 2006

Muilti-dimensional feature vector matching algorithm uses multiple features such as intensity, gradient, variance, first or second derivative of a pixel to find correspondence pixels in stereo images. In this paper, we proposed a new method for adjusting automatically the weight of feature in multi-dimensional feature vector matching considering sharpeness of a pixel in feature vector distance curve. The sharpeness consists of minimum and maximum vector distances of a small window mask. In the experiment we used IKONOS satellite stereo imagery and obtained accurate matching results comparable to the manual weight-adjusting method.