• The Journal of the Acoustical Society of Korea
• /
• v.16 no.2
• /
• pp.71-79
• /
• 1997

In this paper, a codeword-dependent distance normalization(CDDN) and an instar-formed fuzzy smoothing of output distribution are proposed for robust estimation of output probabilities in the FVQ(fuzzy vector quantization)-DHMM(discrete hidden Markov model). The FVQ-DHMM is a variant of DHMM in which the state output probability is estimated by the sum oft he product of the output probability and its weighting factor for each codeword on an input vector. As the performance of the FVQ-DHMM is influenced by weighting factor and output distribution from a state, it is required to get a method to get robust estimation of weighting factors and output distribution for each state. From experimental results, the proposed CDDN method has reduced 24% of error rate over the conventional FVQ-DHMM, and also reduced 79% of error rate when the smoothing of output distribution is also applied to the computation of an output probability. These results indicate that the use of CDDN and the fuzzy smoothing of output distribution to the FVQ-DHMM lead to improved recognition, and therefore it may be used as an alternative to the robust estimation of output probabilities for HMMs.

• The Journal of Korean Institute of Electromagnetic Engineering and Science
• /
• v.23 no.5
• /
• pp.624-631
• /
• 2012

This paper presents a high-efficiency current mode class-D(CMCD) power amplifier for the 13.56 MHz band using a Guanella's 1:1 transmission-line transformer and filtering circuits at the output network. The second and third s are filtered out in the load network of the class-D amplifier. The implemented CMCD power amplifier exhibited a power gain of 13.4 dB and a high power-added efficiency(PAE) of 84.6 % at an output power of 44.4 dBm using the 13.56 MHz CW input signal. The second and third distortion levels were -50.3 dBc and -46.4 dBc at the same output power level, respectively.

• Journal of Biomedical Engineering Research
• /
• v.20 no.1
• /
• pp.53-58
• /
• 1999

Eddy current in MRI systems degrades gradient field linearity and distorts gradient waveform. When the waveform distortion is spatially variant, it is very difficult to perform special imaging techniques such as the echo planar imaging technique or the fast spin echo imaging technique. In this study, we have developed a new technique to estimate the distorted gradient waveforms at any points inside the imaging region using the finite element method. After obtaining the eddy-current-effect transfer function, which represents magnitude and phase characteristics of the gradient field at a particular point, we have used the transfer function to estimate the actual gradient waveforms at the point. To verify the proposed technique, we have compared the estimated gradient waveforms with the measured ones.

• Smart Media Journal
• /
• v.5 no.3
• /
• pp.35-41
• /
• 2016

In this paper, we propose hand-gesture signal recognition based on EPS(Electronic Potential Sensor) using Deep learning model. Extracted signals which from Electronic field based sensor, EPS have much of the noise, so it must remove in pre-processing. After the noise are removed with filter using frequency feature, the signals are reconstructed with dimensional transformation to overcome limit which have just one-dimension feature with voltage value for using convolution operation. Then, the reconstructed signal data is finally classified and recognized using multiple learning layers model based on deep learning. Since the statistical model based on probability is sensitive to initial parameters, the result can change after training in modeling phase. Deep learning model can overcome this problem because of several layers in training phase. In experiment, we used two different deep learning structures, Convolutional neural networks and Recurrent Neural Network and compared with statistical model algorithm with four kinds of gestures. The recognition result of method using convolutional neural network is better than other algorithms in EPS gesture signal recognition.

• The KIPS Transactions:PartB
• /
• v.9B no.5
• /
• pp.653-662
• /
• 2002

Fingerprint recognition is a task to find a matching pattern in a database for a specific persons fingerprint. To accomplish this task, preprocessing, classification, and matching steps are taken for a large-scale fingerprint database but only the matching step is taken without classification for a small-scale database. The primary matching method is based on minutiae (ridge ending point, bifurcation). This matching method, however, requires a very complex computation to extract minutiae and match minutiae-to-minutiae accurately due to translation, rotation, nonlinear deformation of fingerprint and occurrence of spurious minutiae. In addition, this method requires a laborious preprocessing step in order to improve the quality of fingerprint Images. This paper proposes a new simple method to eliminate these problems. With this method, Gabor variance is used instead of minutiae for fingerprint recognition. The Gabor variance is computed from Gabor features that result from filtering a fingerprint image through Gabor filter. In this paper, this method is described and its test result is shown, demonstrating the potential of using this new method for fingerprint recognition.

• Journal of Advanced Navigation Technology
• /
• v.17 no.5
• /
• pp.491-496
• /
• 2013

In this paper, a half-wavelength folded-slot antenna, which can be applied to RF module for 3G mobile communications by which security lights are monitored and controlled, is proposed. The proposed antenna can be regarded as modified folded-slot structure which has the size reduced to a half of conventional ${\lambda}g$ folded-slot antenna and can be placed at the ground plane edge. In spite of that, the proposed antenna still maintain the advantage of conventional folded-slot antenna that input impedance is close to $50{\Omega}$. The antenna is designed and fabricated within the upper space of $40.5{\times}10mm^2$ on $40.5{\times}62mm^2$ substrate for 3G mobile communication frequency band. The measured impedance bandwidth and antenna gain are 390 MHz and 2 dBi respectively.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.12 no.8
• /
• pp.1363-1369
• /
• 2008

As game and character animation industries are growing, techniques for reproducing realistic character behaviors have been required in various fields. Therefore, intensive researches have been performed in order to find various methods for realistic character animation. The most common approaches to character animation involves tedious user input method, simulation with physical laws based on dynamics, and measurement of actors' behaviors with input devices such as motion capture system. These approaches have their own advantages, but they all have common disadvantage in character control. In order to provide users with convenient control, the realistic animation must be generated with high-level parameters, and the modification should also be made with high-level parameters. In this paper we propose techniques for developing an automated character animation tool which operates with high-level parameters, and introduce techniques for developing actual games by utilizing this tool.

• Journal of the Korea Society of Computer and Information
• /
• v.18 no.11
• /
• pp.167-174
• /
• 2013

TIn this paper, we find the face in color images and the ability to track the face was implemented. Face tracking is the work to find face regions in the image using the functions of the computer system and this function is a necessary for the robot. But such as extracting skin color in the image face tracking can not be performed. Because face in image varies according to the condition such as light conditions, facial expressions condition. In this paper, we use the skin color pixel extraction function added lighting compensation function and the entire processing system was implemented, include performing finding the features of eyes, nose, mouth are confirmed as face. Lighting compensation function is a adjusted sine function and although the result is not suitable for human vision, the function showed about 4% improvement. Face features are detected by amplifying, reducing the value and make a comparison between the represented image. The eye and nose position, lips are detected. Face tracking efficiency was good.

• Journal of Aerospace System Engineering
• /
• v.15 no.1
• /
• pp.72-79
• /
• 2021

In this study, as part of the development of an autonomous flying personal aircraft, an equivalent model of the main wing assembly of an Optionally Piloted Personal Air Vehicle (OPPAV) was developed. Reliability of the developed equivalent model was verified by eigenvalue analysis. The main wing assembly consisted of a main wing, an inboard pod, and an outboard pod. First, for developing an equivalent model of each component, components to produce the equivalent model were divided into several sections. Nodes were then created on the axis of the equivalent model at both ends of each section. In addition, static analysis with unit force and unit moment was performed to calculate the deformation or the amount of rotation at the node to be used in the equivalent model. Equivalent axial, bending, and torsional stiffness of each section were calculated by applying the beam theory. Once the equivalent stiffness of each section was calculated, information of a mass and moment of inertia for each section was entered by creating a lumped mass in the center of each section. An equivalent model was developed using beam element. Finally, the reliability of the developed equivalent model was verified by comparison with results of mode analysis of the fine model.

• The Journal of Korean Institute of Information Technology
• /
• v.17 no.5
• /
• pp.13-20
• /
• 2019

Recently, object detection technology based on CNN has been actively studied. Object detection technology is used as an important technology in autonomous vehicles, intelligent image analysis, and so on. In this paper, we propose a rotation change robust object detection system by applying MI-FL (Moment Invariant-Feature Layer) to SSD (Single Shot Multibox Detector) which is one of CNN-based object detectors. First, the features of the input image are extracted based on the VGG network. Then, a total of six feature layers are applied to generate bounding boxes by predicting the location and type of object. We then use the NMS algorithm to get the bounding box that is the most likely object. Once an object bounding box has been determined, the invariant moment feature of the corresponding region is extracted using MI-FL, and stored and learned in advance. In the detection process, it is possible to detect the rotated image more robust than the conventional method by using the previously stored moment invariant feature information. The performance improvement of about 4 ~ 5% was confirmed by comparing SSD with existing SSD and MI-FL.