• The KIPS Transactions:PartB
• /
• v.14B no.4
• /
• pp.287-294
• /
• 2007

Many classification algorithms for real world data suffer from a data class imbalance problem. To solve this problem, various methods have been proposed such as altering the training balance and designing better sampling strategies. The previous methods are not satisfy in the distribution of the input data and the constraint. In this paper, we propose a focused sampling method which is more superior than previous methods. To solve the problem, we must select some useful data set from all training sets. To get useful data set, the proposed method devide the region according to scores which are computed based on the distribution of SOM over the input data. The scores are sorted in ascending order. They represent the distribution or the input data, which may in turn represent the characteristics or the whole data. A new training dataset is obtained by eliminating unuseful data which are located in the region between an upper bound and a lower bound. The proposed method gives a better or at least similar performance compare to classification accuracy of previous approaches. Besides, it also gives several benefits : ratio reduction of class imbalance; size reduction of training sets; prevention of over-fitting. The proposed method has been tested with kNN classifier. An experimental result in ecoli data set shows that this method achieves the precision up to 2.27 times than the other methods.

• The KIPS Transactions:PartB
• /
• v.9B no.5
• /
• pp.571-578
• /
• 2002

In this paper, we propose a video sequence coding scheme called AMV (Auxiliary Motion Vector) to minimize error propagation caused by transmission errors over the Internet. Unlike the conventional coding schemes the AMY coder, for a macroblock in a frame, selects two best matching blocks among several preceding frames. The best matching block, called a primary block, is used for motion compensation of the destination macroblock. The other block, called an auxiliary block, replaces the primary block in case of its loss at the decoder. When a primary block is corrupted or lost during transmission, the decoder can efficiently and simply suppress error propagation to the subsequent frames by replacing the block with an auxiliary block. This scheme has an advantage of reducing both the number and the impact of error propagations. We implemented the proposed coder by modifying H.263 standard coding and evaluated the performance of our proposed scheme in the simulation. The simulation results show that AMV coder is more efficient than the H.263 baseline coder at the high packet loss rate.

• The Journal of Korean Institute of Electromagnetic Engineering and Science
• /
• v.25 no.1
• /
• pp.53-61
• /
• 2014

In this paper, we present the design, fabrication and measurement of high efficiency GaN HEMT power amplifier using harmonic matching technique. In order to achieve high efficiency, harmonic load-pull simulation is performed, that is, the optimum load impedances are determined at $2^{nd}$ and $3^{rd}$ harmonic frequencies as well as at the fundamental. Then, the output matching circuit is designed based on harmonic load-pull simulation. The measurement of the fabricated power amplifier shows the linear gain of 20 dB and $P_{1dB}$(1 dB gain compression point) of 33.7 dBm at 1.85 GHz. The maximum power added efficiency(PAE) of 80.9 % is achieved at the output power of 38.6 dBm, which belongs to best efficiency performance among the reported high efficiency power amplifiers. For W-CDMA input signal, the power amplifier shows a PAE of 27.8 % at the average output power of 28.4 dBm, where an ACLR (Adjacent Channel Leakage Ratio) is measured to be -38.8 dBc. Digital predistortion using polynomial fitting was implemented to linearize the power amplifiers, which allowed about 6.2 dB improvement of an ACLR performance.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.38 no.2
• /
• pp.117-128
• /
• 2001

In order to achieve optimal retrieval, i.e., to find the best match to a query according to a certain similarity measure, the exhaustive search should be performed literally for all the images in a database. However, the straightforward exhaustive search algorithm is computationally expensive in large image databases. To reduce its heavy computational cost, this paper presents a fast exhaustive multi-resolution search algorithm based on image database clustering. Firstly, the proposed algorithm partitions the whole image data set into a pre-defined number of clusters having similar feature contents. Next, for a given query, it checks the lower bound of distances in each cluster, eliminating disqualified clusters. Then, it only examines the candidates in the remaining clusters. To alleviate unnecessary feature matching operations in the search procedure, the distance inequality property is employed based on a multi-resolution data structure. The proposed algorithm realizes a fast exhaustive multi-resolution search for either the best match or multiple best matches to the query. Using luminance histograms as a feature, we prove that the proposed algorithm guarantees optimal retrieval with high searching speed.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.4 no.1
• /
• pp.25-44
• /
• 2010

With the increasing popularity of mobile devices, it has become necessary to protect private information and content in these devices. Face recognition has been favored over conventional passwords or security keys, because it can be easily implemented using a built-in camera, while providing user convenience. However, because mobile devices can be used both indoors and outdoors, there can be many illumination changes, which can reduce the accuracy of face recognition. Therefore, we propose a new face recognition method on a mobile device robust to illumination variations. This research makes the following four original contributions. First, we compared the performance of face recognition with illumination variations on mobile devices for several illumination normalization procedures suitable for mobile devices with low processing power. These include the Retinex filter, histogram equalization and histogram stretching. Second, we compared the performance for global and local methods of face recognition such as PCA (Principal Component Analysis), LNMF (Local Non-negative Matrix Factorization) and LBP (Local Binary Pattern) using an integer-based kernel suitable for mobile devices having low processing power. Third, the characteristics of each method according to the illumination va iations are analyzed. Fourth, we use two matching scores for several methods of illumination normalization, Retinex and histogram stretching, which show the best and $2^{nd}$ best performances, respectively. These are used as the inputs of an SVM (Support Vector Machine) classifier, which can increase the accuracy of face recognition. Experimental results with two databases (data collected by a mobile device and the AR database) showed that the accuracy of face recognition achieved by the proposed method was superior to that of other methods.

• JSTS:Journal of Semiconductor Technology and Science
• /
• v.15 no.2
• /
• pp.301-306
• /
• 2015

An ultra-compact and wideband low noise amplifier (LNA) in a quad flat non-leaded (QFN) package is presented. The LNA monolithic microwave integrated circuit (MMIC) is implemented in a $0.25{\mu}m$ GaN IC technology on a Silicon Carbide (SiC) substrate provided by Triquint. A source degeneration inductor and a gate inductor are used to obtain the noise and input matching simultaneously. The resistive feedback and inductor peaking techniques are employed to achieve a wideband characteristic. The LNA chip is mounted in the $3{\times}3-mm^2$ QFN package and measured. The supply voltages for the first and second stages are 14 V and 7 V, respectively, and the total current is 70 mA. The highest gain is 13.5 dB around the mid-band, and -3 dB frequencies are observed at 0.7 and 12 GHz. Input and output return losses ($S_{11}$ and $S_{22}$) of less than -10 dB measure from 1 to 12 GHz; there is an absolute bandwidth of 11 GHz and a fractional bandwidth of 169%. Across the bandwidth, the noise figures (NFs) are between 3 and 5 dB, while the output-referred third-order intercept points (OIP3s) are between 26 and 28 dBm. The overall chip size with all bonding pads is $1.1{\times}0.9mm^2$. To the best of our knowledge, this LNA shows the best figure-of-merit (FoM) compared with other published GaN LNAs with the same gate length.

• The Journal of the Korea Contents Association
• /
• v.11 no.2
• /
• pp.31-39
• /
• 2011

The facial recognition algorithms using Gabor wavelet based face graph produce very good performance while they have some weakness such as a large amount of computation and an irregular result depend on initial location. We proposed a fully automatic facial recognition algorithm using a Gabor feature based geometric deformable face graph matching. The initial location and size of a face graph can be selected using Adaboost detection results for speed-up. To find the best face graph with the face model graph by updating the size and location of the graph, the geometric transformable parameters are defined. The best parameters for an optimal face graph are derived using an optimization technique. The simulation results show that the proposed algorithm can produce very good performance with recognition rate 96.7% and recognition speed 0.26 sec for FERET database.

• Journal of Korea Multimedia Society
• /
• v.20 no.7
• /
• pp.979-985
• /
• 2017

This paper, which covers a fluorescence microscopy image of brain tumor cells, looks at drug reactions by treating different types and concentrations of drugs on a plate of $24{\times}16$ wells. Due to the limitation of the field of view, a well was taken into 9 field images, and each has an overlapping area with its neighboring fields. To analyze more precisely, image stitching is needed. The basic method is finding a similar area using normalized cross-correlation (NCC). The problem is that some overlapping areas may not have any duplicated cells that help to find the matching point. In addition, the cell objects have similar sizes and shapes, which makes distinguishing them difficult. To avoid calculating similarity between blank areas and roughly distinguishing different cells, thresholding is added. The thresholding method classifies background and cell objects based on fixed thresholds and finds the location of the first seen cell. After getting its location, NCC is used to find the best correlation point. The results are compared with a simple boundary stitched image. Our proposed method stitches images that are connected in a grid form without collision, selecting the best correlation point among areas that contain overlapping cells and ones without it.

• Journal of Microbiology and Biotechnology
• /
• v.28 no.10
• /
• pp.1671-1682
• /
• 2018

Alginate lyases (endo and exo-lyases) are required for the degradation of alginate into its constituting monomers. Efficient bioethanol production and extraction of bioactives from brown algae requires intensive use of these enzymes. Nonetheless, there are few commercial alginate lyase preparations, and their costs make them unsuitable for large scale experiments. A recombinant expression protocol has been developed in this study for producing seven endo-lyases and three exo-lyases as soluble and highly active preparations. Saccharification of alginate using 21 different endo/exo-lyase combinations shows that there is complementary enzymatic activity between some of the endo/exo pairs. This is probably due to favorable matching of their substrate biases for the different glycosidic bonds in the alginate molecule. Therefore, selection of enzymes for the best saccharification results for a given biomass should be based on screens comprising both types of lyases. Additionally, different incubation temperatures, enzyme load ratios, and enzyme loading strategies were assessed using the best four enzyme combinations for treating Macrocystis pyrifera biomass. It was shown that $30^{\circ}C$ with a 1:3 endo/exo loading ratio was suitable for all four combinations. Moreover, simultaneous loading of endo-and exo-lyases at the beginning of the reaction allowed maximum alginate saccharification in half the time than when the exo-lyases were added sequentially.

• Journal of Digital Contents Society
• /
• v.18 no.2
• /
• pp.347-356
• /
• 2017

It is well-known that the face recognition method via sparse representation has been proved very robust and showed good performance. Its weakness is, however, that its time complexity is very high because it should solve $L_1$-minimization problem to find the sparse solution. In this paper, we propose to use the ROMP(Regularized Orthogonal Matching Pursuit) method for the sparse solution, which solves the $L_2$-minimization problem with regularization condition using the greed strategy. In experiments, we shows that the proposed method is comparable to the existing best $L_1$-minimization solver, Homotopy, but is 60 times faster than Homotopy. Also, we proposed C-SCI method for classification. The C-SCI method is very effective since it considers the sparse solution only without reconstructing the test data. It is shown that the C-SCI method is comparable to, but is 5 times faster than the existing best classification method.