• Speech Sciences
• /
• v.10 no.4
• /
• pp.159-166
• /
• 2003

To reduce the high dimensionality required for training of feature vectors in speaker identification, we propose an efficient GMM based on local PCA with Fuzzy clustering. The proposed method firstly partitions the data space into several disjoint clusters by fuzzy clustering, and then performs PCA using the fuzzy covariance matrix in each cluster. Finally, the GMM for speaker is obtained from the transformed feature vectors with reduced dimension in each cluster. Compared to the conventional GMM with diagonal covariance matrix, the proposed method needs less storage and shows faster result, under the same performance.

• The Journal of the Korea Contents Association
• /
• v.9 no.4
• /
• pp.121-130
• /
• 2009

In this paper, feature extraction methods, which is one field of reducing dimensions of high-dimensional data, are empirically investigated. We selected the traditional PCA(Principal Component Analysis), ICA(Independent Component Analysis), NMF(Non-negative Matrix Factorization), and sNMF(Sparse NMF) for comparisons. ICA has a similar feature with the simple cell of V1. NMF implemented a "parts-based representation in the brain" and sNMF is a improved version of NMF. In order to visually investigate the extracted features, handwritten digits are handled. Also, the extracted features are used to train multi-layer perceptrons for recognition test. The characteristic of each feature extraction method will be useful when applying feature extraction methods to many real-world problems.

• Korean Journal of Remote Sensing
• /
• v.30 no.6
• /
• pp.769-776
• /
• 2014

It is necessary to detect the feature points existing simultaneously in both images and then find the corresponding relationship between the detected feature points. We propose a new feature detector based on geometric mean of two eigenvalues of gradient matrix which is able to measure the change of pixel intensities. The corner response of the proposed detector is proportional to the geometric mean and also the difference of two eigenvalues in the case of same geometric mean. We analyzed the localization error of the feature detection using aerial image and artificial image with various types of corners. The localization error of the proposed detector was smaller than that of the typical corner detector, Harris detector.

• Proceedings of the KSRS Conference
• /
• 1999.11a
• /
• pp.8-12
• /
• 1999

Since the advent of high resolution satellite image, possibilities of applying various human interpretation mechanism to these images have increased. Also many studies about these possibilities in many fields such as computer vision, pattern recognition, artificial intellegence and remote sensing have been done. In this field of these studies, texture is defined as a kind of quantity related to spatial distribution of brightness and tone and also plays an important role for interpretation of images. Especially, methods of obtaining texture by statistical model have been studied intensively. Among these methods, texture measurement method based on cooccurrence matrix is highly estimated because it is easy to calculate texture features compared with other methods. In addition, these results in high classification accuracy when this is applied to satellite images and aerial photos. But in the existing studies using cooccurrence matrix, features have been chosen arbitrarily without considering feature variation. And not enough studies have been implemented for appropriate resolution selection in which cooccurrence matrix can extract texture. Therefore, this study reviews the concept of cooccurrence matrix as a texture measurement method, evaluates usefulness of several features obtained from cooccurrence matrix, and proposes appropriate resolution by investigating variance trend of several features.

• Journal of Korea Multimedia Society
• /
• v.20 no.1
• /
• pp.43-50
• /
• 2017

According to living environment, the number of deaths due to thyroid diseases increased. In this paper, we proposed an algorithm for recognizing a thyroid detection using texture analysis based on shape, gray level co-occurrence matrix and gray level run length matrix. First of all, we segmented the region of interest (ROI) using active contour model algorithm. Then, we applied a total of 18 features (5 first order descriptors, 10 Gray level co-occurrence matrix features(GLCM), 2 Gray level run length matrix features and shape feature) to each thyroid region of interest. The extracted features are used as statistical analysis. Our results show that first order statistics (Skewness, Entropy, Energy, Smoothness), GLCM (Correlation, Contrast, Energy, Entropy, Difference variance, Difference Entropy, Homogeneity, Maximum Probability, Sum average, Sum entropy), GLRLM features and shape feature helped to distinguish thyroid benign and malignant. This algorithm will be helpful to diagnose of thyroid nodule on ultrasound images.

• The KIPS Transactions:PartB
• /
• v.13B no.4 s.107
• /
• pp.489-498
• /
• 2006

Clustering results differ according to the datasets and the performance worsens even while using web documents which are manually processed by an indexer, because although representative clusters for a feature can be obtained by statistical feature selection methods, irrelevant features(i.e., non-obvious features and those appearing in general documents) are not eliminated. Those irrelevant features should be eliminated for improving clustering performance. Therefore, this paper proposes three feature-filtering algorithms which consider feature values per document set, together with distribution, frequency, and weights of features per document set: (l) features filtering algorithm in a document (FFID), (2) features filtering algorithm in a document matrix (FFIM), and (3) a hybrid method combining both FFID and FFIM (HFF). We have tested the clustering performance by feature selection using term frequency and expand co link information, and by feature filtering using the above methods FFID, FFIM, HFF methods. According to the results of our experiments, HFF had the best performance, whereas FFIM performed better than FFID.

• Journal of Institute of Control, Robotics and Systems
• /
• v.22 no.6
• /
• pp.403-410
• /
• 2016

This paper proposes a robust image-based visual servoing scheme using a nonlinear observer for a monocular eye-in-hand manipulator. The proposed control method is divided into a range estimation phase and a target-tracking phase. In the range estimation phase, the range from the camera to the target is estimated under the non-moving target condition to solve the uncertainty of an interaction matrix. Then, in the target-tracking phase, the feature point uncertainty caused by the unknown motion of the target is estimated and feature point errors converge sufficiently near to zero through compensation for the feature point uncertainty.

• Journal of the Korea Institute of Military Science and Technology
• /
• v.22 no.2
• /
• pp.207-214
• /
• 2019

This paper proposes a new feature extraction method for automatically estimating the number of target and detecting the chaff using high range resolution profile(HRRP). Feature of one-dimensional range profile is expected to be limited or missing due to lack of information according to the time. The proposed method considers the dynamic movements of targets depending on the radial velocity. The observed HRRP sequence is used to construct a time-range distribution matrix, then assuming diverse radial velocities reflect the number of target and seduction chaff launch, the proposed method utilizes the characteristic of the gradient distribution on the time-range distribution matrix image, which is validated by electromagnetic computation data and dynamic simulation.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.17 no.12
• /
• pp.1437-1450
• /
• 1992

This paper describes the feature extraction of the interdigital regions of palm, and proposes a personal verification algorithm using the extracted features and the pattern types of those. The procedures of the feature extraction are as follows : first, the interdigital region is partitioned into several subregions, examining the phase of rigdes in each subregion, deciding the direction of that phase, and making the direction matrix of the region, we analyze this direction matrix to contain a feature pattern, and then, yield the first core. Second, applying the thinning to around the first core and tracing the thinned ridges, we yield the feature pattern types and second cores. Finally, the feature patterns coordinates included all of them are built. Then, distances and directions from each second core reaching to all the others are yielded from that coordinates. These informations are used to make a feature parameter. In our verification algorithm, such pattern types, the numbers of feature patterns, theses positions and feature parameters are used to analyze.

• Convergence Security Journal
• /
• v.8 no.4
• /
• pp.161-166
• /
• 2008

This paper proposes a new approach to align 3D data sets by using curvatures of feature surface. We use the Gaussian curvatures and the covariance matrix which imply the physical characteristics of the model to achieve registration of unaligned 3D data sets. First, the physical characteristics of local area are obtained by the Gaussian curvature. And the camera position of 3D range finder system is calculated from by using the projection matrix between 3D data set and 2D image. Then, the physical characteristics of whole area are obtained by the covariance matrix of the model. The corresponding points can be found in the overlapping region with the cross-projection method and it concentrates by removed points of self-occlusion. By the repeatedly the process discussed above, we finally find corrected points of overlapping region and get the optimized registration result.