• Geophysics and Geophysical Exploration
• /
• v.1 no.3
• /
• pp.176-182
• /
• 1998

The borehole radar reflection survey can image the underground structure with high resolution, however, we cannot get any information on the orientation of the reflectors with dipole antenna alone. The direction-finding antenna system is commonly used to give the solution to the problem. However, the interpretation of the data from direction- finding antenna may be time-consuming, and sometimes have ambiguities in the sense of precise determination of the azimuth. To solve the problem, we developed the automatic azimuth finding scheme of reflectors in borehole radar reflection data using direction-finding antenna. The algorithm is based on finding the azimuthal angle possibly showing the maximum reflection amplitude in the least-squared error sense. The developed algorithm was applied to the field data acquired in quarry mine. It was possible to locate nearly all of the reflectors in three dimensional fashion, which coincide with the known geological structures and man-made discontinuities.

• The Journal of the Acoustical Society of Korea
• /
• v.39 no.5
• /
• pp.497-504
• /
• 2020

Constructing speaker embeddings in speaker verification is an important issue. In general, a self-attention mechanism has been applied for speaker embedding encoding. Previous studies focused on training the self-attention in a high-level layer, such as the last pooling layer. In this case, the effect of low-level layers is not well represented in the speaker embedding encoding. In this study, we propose Masked Cross Self-Attentive Encoding (MCSAE) using ResNet. It focuses on training the features of both high-level and low-level layers. Based on multi-layer aggregation, the output features of each residual layer are used for the MCSAE. In the MCSAE, the interdependence of each input features is trained by cross self-attention module. A random masking regularization module is also applied to prevent overfitting problem. The MCSAE enhances the weight of frames representing the speaker information. Then, the output features are concatenated and encoded in the speaker embedding. Therefore, a more informative speaker embedding is encoded by using the MCSAE. The experimental results showed an equal error rate of 2.63 % using the VoxCeleb1 evaluation dataset. It improved performance compared with the previous self-attentive encoding and state-of-the-art methods.

• Journal of Wetlands Research
• /
• v.16 no.3
• /
• pp.441-450
• /
• 2014

In this study, we consider ability of self-purification for a rational water quality management. And we assess the risk of Alkyl Benzene Sulfonic acid sodium salt(ABS) of harmful ingredients in Anseong Cheon watershed using QUAL2E model. The observations and simulated results were fitted well for BOD and ABS, but even though the trend of DO concentration change was well represented, the error between observation and simulation values was existed. We assessed the Risk assessment by calculating Risk quotient(RQ) by Predicted Exposure Concentration(PEC) and Predicted No-Effect Concentration(PNEC). Results of the impact of ABS on the self-purification of the river were Anseongcheon[0.0003(Bressan), 0.06(Criteria of Ministry of environment)], Jinwicheon[0.0002(Bressan), 0.04(Criteria of Ministry of environment). And result of the impact of ABS on the Aquatic ecosystem of the river were Anseongcheon[0.0667(Bressan), 0.005(Criteria of Ministry of environment)], Jinwicheon[0.1(Bressan), 0.0075(Criteria of Ministry of environment). All of these results were smaller than the 1.0 which is the reference value suggested by Norification No.30 of the National Institute of Environment Research. So, ABS did not affect a self-purification and aquatic ecosystem of the river. The method suggested in the study is a simple one and can provide more information for harmful ingredients than criteria of Ministry of environment.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.16 no.2
• /
• pp.101-106
• /
• 2016

This paper proposes the SU-SE-MMA algorithm which applying the concept of selective updaing to the SE-MMA that is possible to reduce the intersymbol interference due to distortion occurred at the channel when transmit the nonconstant modulus 16-QAM signal. The SE-MMA emerged for the simplifying the computational operation from the current MMA adaptation algorithm, then it's has the fast convergence speed and has a problem of increase the residual component in the steady state. The SU-SE-MMA performs the selectively tap updating when the distance of equalizer output and specified transmit signal point is greater than the given threshold value and tap updaing does not occurred in the small distance. By this selective updating process, it is possible to more reduction in the computational operation in the propose algorithm. The improved adaptive equalization performance of SU-SE-MMA like as the equalizer output signal constellation, residual isi, MD, SER were confirmed by computer simulation compared to SE-MMA. As a result of simulation, the AV-SE-MMA has better performance in output signal constellation, residual isi and MD compared to the SE-MMA, but it was confirmed that the AV-SE-MMA has similar in the SER performance that means the robustness to the noise.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.15 no.3
• /
• pp.99-105
• /
• 2015

This paper deals with the mMMA (modified MMA) which possible to improving the convergence speed that employing the serial connecting form of two stage digital filter instead of signal filter of MMA adaptive equalizer without applying the variable step size for compensates the intersymbol interference by channel distortion in the nonconstant modulus signal. The adaptive equalizer can be implemented by signal digital filter using the finite order tap delay line. In this paper, the equalizer is implemented by the two stage serial form and the filter coefficient are updated by the error signal using the same algorithm of MMA in each stage. The fast convergence speed is determined in the first stage, and the residual isi left at the output of first stage output is minimized in the second stage filter. The same digital filter length was considered in single stage and two stage system and the performance of these systems were compared. The performance index includes the output signal constellation, the residual isi and maximum distortion, MSE that is measure of the convergence characteristics, the SER. As a result of computer simulation, mMMA that has a FIR structure of two stage, has more good performance in every performance index except the constellation diagram due to equalization noise and improves the convergence speed about 1.5~1.8 time than the present MMA that has a FIR structure of single stage.

• Journal of the Korean Institute of Telematics and Electronics T
• /
• v.35T no.1
• /
• pp.90-98
• /
• 1998

This paper is to describe the improved vector control which can control the induction motor robustly in low speed. When the induction motor is drived with low speed, below 10 percent of the rated speed, an algorithm which can compensate the error of unit vector angle generated by the harmonics is proposed. Another algorithm which can be tuned to the rotor time constant so that nay be robust to the rotor parameter change in low speed and transient state was proposed. The ripple of flux and torque was reduced by the proposed vector control and then the stable output characteristics was obtained in low speed. When the input and output is sinusoidal, the proposed vector control, the direct vector control and the indirect vector control were analyzed and compared in the low speed characteristics. And each control characteristics is compared and analyzed in state of containing harmonics. The estimation and tunning performance of rotor time constant is confirmed with simulation. The whole control system is implemented by real hardware and experimented to compare the proposed vector control with the direct vector control. As a result of the experiment with two control methods in low speed, the torque ripple of the proposed vector control is improved by 45 percent than the direct vector control. And it is confirmed that the flux current ripple is reduced in 0.2 p.u. and torque current ripple is reduced in 0.6 p.u. It is confirmed that the rotor time constant by the estimation and the tunning algorithm is tunned by the real rotor time constant. Finally, it was confirmed that the validity and robustness for the proposed vector control in low speed existed.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.41 no.2
• /
• pp.79-88
• /
• 2004

Gaze detection is to locate the position on a monitor screen where a user is looking by computer vision. Gaze detection systems have numerous fields of application. They are applicable to the man-machine interface for helping the handicapped to use computers and the view control in three dimensional simulation programs. In our work, we implement it with a computer vision system setting a IR-LED based single camera. To detect the gaze position, we locate facial features, which is effectively performed with IR-LED based camera and SVM(Support Vector Machine). When a user gazes at a position of monitor, we can compute the 3D positions of those features based on 3D rotation and translation estimation and affine transform. Finally, the gaze position by the facial movements is computed from the normal vector of the plane determined by those computed 3D positions of features. In addition, we use a trained neural network to detect the gaze position by eye's movement. As experimental results, we can obtain the facial and eye gaze position on a monitor and the gaze position accuracy between the computed positions and the real ones is about 4.8 cm of RMS error.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.43 no.4 s.310
• /
• pp.74-86
• /
• 2006

This paper proposes a robust and computationally efficient algorithm for automatic video object segmentation. For implementing the spatio-temporal segmentation, which aims for efficient combination of the motion segmentation and the color segmentation, an SOM-based hierarchical clustering method in which the segmentation process is regarded as clustering of feature vectors is employed. As results, problems of high computational complexity which required for obtaining exact segmentation results in conventional video object segmentation methods, and the performance degradation due to noise are significantly reduced. A measure of motion vector reliability which employs MRF-based MAP estimation scheme has been introduced to minimize the influence from the motion estimation error. In addition, a noise elimination scheme based on the motion reliability histogram and a clustering validity index for automatically identifying the number of objects in the scene have been applied. A cross projection method for effective object tracking and a dynamic memory to maintain temporal coherency have been introduced as well. A set of experiments has been conducted over several video sequences to evaluate the proposed algorithm, and the efficiency in terms of computational complexity, robustness from noise, and higher segmentation accuracy of the proposed algorithm have been proved.

• Journal of the Institute of Electronics Engineers of Korea SC
• /
• v.47 no.4
• /
• pp.30-39
• /
• 2010

Since Diffusion tensor from Diffusion Tensor Magnetic Resonance Imaging(DT-MRI) is so sensitive to noise, the principle eigenvector(PEV) calculated from Diffusion tensor could be erroneous. Tractography obtained from PEV could be deviated from the real fiber tract. Therefore regularization process is needed to eliminate noise. In this paper, to reduce noise in DT-MRI measurements, the Dyadic Sorting(DS) method as regularization of the eigenvalue and the eigenvector is applied in the tractography. To resort the eigenvalues and the eignevectors, the DS method uses the intervoxel overlap function which can measure the overlap between eigenvalue-eigenvector pairs in the $3\times3$ pixel. In this paper, we applied the DS method to the three-dimensional volume. We discuss the error analysis and numerical study to the synthetic and the experimental data. As a result, we have shown that the DS method is more efficient than the median filtering methods as much as 79.97%~83.64%, 85.62%~87.76% in AAE, AFA respectively for the corticospinal tract of the experimental data.

• Journal of the Institute of Electronics Engineers of Korea SC
• /
• v.49 no.1
• /
• pp.29-38
• /
• 2012

The noise generated in the process of obtaining the medical image acts as the element obstructing the image interpretation and diagnosis. To restore the true image from the image polluted from the noise, the total variation optimization algorithm was proposed by the R.O. F (L.Rudin, S Osher, E. Fatemi). This method removes the noise by fitting the balance of the regularity and fidelity. However, the blurring phenomenon of the border area generated in the process of performing the iterative operation cannot be avoided. In this paper, we propose the adaptive total variation method by mapping the control parameter to the proposed transfer function for minimizing boundary error. The proposed transfer function is determined by the noise variance and the local property of the image. The proposed method was applied to 464 tooth images. To evaluate proposed method performance, PSNR which is a indicator of signal and noise's signal power ratio was used. The experimental results show that the proposed method has better performance than other methods.