• The Journal of the Acoustical Society of Korea
• /
• v.41 no.3
• /
• pp.326-334
• /
• 2022

In this paper, we propose speech enhancement algorithm as a pre-processing for robust speech recognition in noisy environments. Auxiliary-function-based Independent Vector Analysis (AuxIVA) is performed with weighted covariance matrix using time-varying variances with scaling factor from target masks representing time-frequency contributions of target speech. The mask estimates can be obtained using Neural Network (NN) pre-trained for speech extraction or diffuseness using Coherence-to-Diffuse power Ratio (CDR) to find the direct sounds component of a target speech. In addition, outputs for omni-directional noise are closely chained by sharing the time-varying variances similarly to independent subspace analysis or IVA. The speech extraction method based on AuxIVA is also performed in Independent Low-Rank Matrix Analysis (ILRMA) framework by extending the Non-negative Matrix Factorization (NMF) for noise outputs to Non-negative Tensor Factorization (NTF) to maintain the inter-channel dependency in noise output channels. Experimental results on the CHiME-4 datasets demonstrate the effectiveness of the presented algorithms.

• Composites Research
• /
• v.36 no.1
• /
• pp.16-24
• /
• 2023

This paper presents a multi-scale progressive fatigue damage model incorporating the model for interfacial debonding between fibers and matrix. The micromechanics model for the progressive interface debonding was adopted, which defined the four different interface phases: (1) perfectly bonded fibers; (2) mild imperfect interface; (3) severe imperfect interface; and (4) completely debonded fibers. As the number of cycles increases, the progressive transition from the perfectly bonded state to the completely debonded fiber state occurs. Eshelby's tensor for each imperfect state is calculated by the linear spring model for a damaged interface, and effective elastic properties are obtained using the multi-phase homogenization method. The fatigue damage evolution formulas for fiber, matrix and interface were proposed to demonstrate the fatigue behavior of CFRP laminates under cyclic loading. The material parameters for the fiber/matrix fatigue damage were characterized using the chaotic firefly algorithm. The model was implemented into the UMAT subroutine of ABAQUS, and successfully validated with flat-bar UD laminate specimens ([0]₈,[90]₈, [30]₁₆) of AS4/3501-6 graphite/epoxy composite.

• Composites Research
• /
• v.36 no.3
• /
• pp.205-210
• /
• 2023

In this study, we developed a transfer learning framework based on homogenization data for efficient prediction of the effective mechanical properties and thermal conductivity of cellular foam structures. Mean-field homogenization (MFH) based on the Eshelby's tensor allows for efficient prediction of properties in porous structures including ellipsoidal inclusions, but accurately predicting the properties of cellular foam structures is challenging. On the other hand, finite element homogenization (FEH) is more accurate but comes with relatively high computational cost. In this paper, we propose a data-driven transfer learning framework that combines the advantages of mean-field homogenization and finite element homogenization. Specifically, we generate a large amount of mean-field homogenization data to build a pre-trained model, and then fine-tune it using a relatively small amount of finite element homogenization data. Numerical examples were conducted to validate the proposed framework and verify the accuracy of the analysis. The results of this study are expected to be applicable to the analysis of materials with various foam structures.

• Journal of Internet of Things and Convergence
• /
• v.10 no.2
• /
• pp.1-15
• /
• 2024

The construction of smart communities is a new method and important measure to ensure the security of residential areas. In order to solve the problem of low accuracy in face recognition caused by distorting facial features due to monitoring camera angles and other external factors, this paper proposes the following optimization strategies in designing a face recognition network: firstly, a global graph convolution module is designed to encode facial features as graph nodes, and a multi-scale feature enhancement residual module is designed to extract facial keypoint features in conjunction with the global graph convolution module. Secondly, after obtaining facial keypoints, they are constructed as a directed graph structure, and graph attention mechanisms are used to enhance the representation power of graph features. Finally, tensor computations are performed on the graph features of two faces, and the aggregated features are extracted and discriminated by a fully connected layer to determine whether the individuals' identities are the same. Through various experimental tests, the network designed in this paper achieves an AUC index of 85.65% for facial keypoint localization on the 300W public dataset and 88.92% on a self-built dataset. In terms of face recognition accuracy, the proposed network achieves an accuracy of 83.41% on the IBUG public dataset and 96.74% on a self-built dataset. Experimental results demonstrate that the network designed in this paper exhibits high detection and recognition accuracy for faces in surveillance videos.

• Journal of Service Research and Studies
• /
• v.14 no.2
• /
• pp.18-36
• /
• 2024

As interest in ESG has been increased, it is easy to find papers that empirically study that a company's ESG activities have a positive impact on the company's performance. However, research on what ESG activities companies should actually engage in is relatively lacking. Accordingly, this study systematically classifies ESG activities of companies and seeks to provide insight to companies seeking to plan new ESG activities. This study analyzes how Chinese manufacturing companies perform ESG activities based on their dynamic capabilities in the global economy and how they differ in their activities. This study used the ESG annual reports of 151 Chinese manufacturing listed companies on the Shanghai & Shenzhen Stock Exchange and ESG indicators of China Securities Index Company (CSI) as data. This study focused on the following three research questions. The first is to determine whether there are any differences in ESG activities between companies with high ESG scores (TOP-25) and companies with low ESG scores (BOT-25), and the second is to determine whether there are any changes in ESG activities over a 10-year period (2010-2019), focusing only on companies with high ESG scores. The results showed that there was a significant difference in ESG activities between high and low ESG scorers, while tracking the year-to-year change in activities of the top-25 companies did not show any difference in ESG activities. In the third study, social network analysis was conducted on the keywords of E/S/G. Through the co-concurrence matrix technique, we visualized the ESG activities of companies in a four-quadrant graph and set the direction for ESG activities based on this.

• Investigative Magnetic Resonance Imaging
• /
• v.4 no.1
• /
• pp.27-33
• /
• 2000

Purpose: To determine the electronic spin relaxation times, $T_{le}$, of three commercially available Gd-chelated MR contrast agents, Gd-DTPA, Gd-DTPA-BMA and Gd-DOTA, using Electron Paramagnetic Resonance(EPR) technique. Material and Methods: The paramagnetic MR contrast agents, Gd-DTFA(Magnevist) , Gd-DTFA-BMA(OMNISCAN) and Gd-DOTA(Dotarem), were used for this study, The EPR spectra of these contrast agents, which were prepared 2:1 methanol/water solution, were obtained at low temperatures, from $-160^{\circ}C~20^{\circ}C$. The glassy-state EPR spectra for these contrast agents were then fitted by the simulation spectra generated with different zero-field splitting (ZFS) parameters by a computer simulation program 'GEN', which generates the EPR powder spectrum using a given ZFS in $3{\times}3$ tensor. Finally, the spin relaxation times of the contrast agents were then determined from the $T_{2e}$, D, and E values of the best simulation spectra using the McLachlan's theory of average relaxation rate. Results: The electronic transverse spin relaxation times, $T_{2e}'s$, of Gd-DTPA, Gd-DTPA-BMA and Gd-DOTA were 0.113ns, 0.147ns and 1.81ns respectively. The g-values were 1.9737, 1.9735 and 1.9830 and the electronic spin relaxation times, $T_{1e}'s$, were 18.70ns, 33.40ns and $1.66{\mu}s$, respectively. Conclusion: The results of these studies reconfirm that the paramagnetic MR contrast agents with larger ZFS parameters should have shorter $T_{1e}'s$. Among three contrast agents used for this study, Gd-DOTA chelated with cyclic ligand structure shows better electronic property then the others with linear structure. Thus, it is concluded that the exact determination of ZFS parameters is the important factor in evaluating relaxation enhancement effect of the agents and in developing new contrast agents.

• Economic and Environmental Geology
• /
• v.39 no.2 s.177
• /
• pp.111-128
• /
• 2006

Three dimensional crustal structure and source features of earthquake hypocenters on the Korean peninsula were investigated using P and S-wave travel time tomography. The main goal of this research was to find Vp/Vs anomalies at earthquake hypocenters as well as those of crustal structure of basins and deep tectonic settings. This allowed fer the extrapolation of more detailed seismotectonic force from the Korean peninsula. The earthquake hypocenters were found to have high Vp/Vs ratio discrepancies (VRD) at the vertical sections. High V/p/Vs ratios were also found in the sedimentary basins and beneath the Chugaryong Rift Zone (CRZ), which was due to mantle plume that subsequently solidified with many fractures and faults which were saturated with connate water. The hypocenters of most earthquakes were found in the upper crust for Youngwol (YE), Kyongju (KE), Hongsung (HE), Kaesong (KSE), Daekwan (DKE), and Daehung (DHE) earthquakes, but near the subcrust or the Moho Discontinuity for Mt. Songni (SE), Sariwon (SRE) and Mt. Jiri (JE) earthquakes. Especially, we found hot springs of the Daekwan, Daehung and Unsan regions coincide with high VRD. Also, this cannot rule out the possibility that there are some partial meltings in the subcrust of this region. High VRD might indicate that many faults and fractures with connate water were dehydrated when earthquakes took place, reducing shear modulus in the hypocenter areas. This is can be explained by due to the fact that a point source which is represented by the moment tensor that may involve changes in volume, shear fracture, and rigidity. High Vp/Vs ratio discrepancies (VRD) were also found beneath Mt. Backdu beneath 40 km, indicating that magma chamber existed beneath Mt. Backdu is reducing shear modulus of S-wave velocity.