• Phonetics and Speech Sciences
• /
• v.10 no.1
• /
• pp.33-38
• /
• 2018

Although deep neural network-based acoustic models have greatly improved the performance of automatic speech recognition (ASR), reverberation still degrades the performance of distant speech recognition in indoor environments. In this paper, we adopt the DenseNet, which has shown great performance results in image classification tasks, to improve the performance of reverberant speech recognition. The DenseNet enables the deep convolutional neural network (CNN) to be effectively trained by concatenating feature maps in each convolutional layer. In addition, we extend the concept of multi-resolution CNN to multi-resolution DenseNet for robust speech recognition in reverberant environments. We evaluate the performance of reverberant speech recognition on the single-channel ASR task in reverberant voice enhancement and recognition benchmark (REVERB) challenge 2014. According to the experimental results, the DenseNet-based acoustic models show better performance than do the conventional CNN-based ones, and the multi-resolution DenseNet provides additional performance improvement.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.12 no.6
• /
• pp.2595-2618
• /
• 2018

In this paper, we investigate secure communication with the presence of multiple eavesdroppers (Eves) in a two-tier downlink dense heterogeneous network, wherein there is a macrocell base station (MBS) and multiple femtocell base stations (FBSs). Each base station (BS) has multiple users. And Eves attempt to wiretap a macrocell user (MU). To keep Eves ignorant of the confidential message, we propose a physical-layer security scheme based on cross-layer cooperation to exploit interference in the considered network. Under the constraints on the quality of service (QoS) of other legitimate users and transmit power, the secrecy rate of system can be maximized through jointly optimizing the beamforming vectors of MBS and cooperative FBSs. We explore the problem of maximizing secrecy rate in both non-colluding and colluding Eves scenarios, respectively. Firstly, in non-colluding Eves scenario, we approximate the original non-convex problem into a few semi-definite programs (SDPs) by employing the semi-definite relaxation (SDR) technique and conservative convex approximation under perfect channel state information (CSI) case. Furthermore, we extend the frame to imperfect CSI case and use the Lagrangian dual theory to cope with uncertain constraints on CSI. Secondly, in colluding Eves scenario, we transform the original problem into a two-tier optimization problem equivalently. Among them, the outer layer problem is a single variable optimization problem and can be solved by one-dimensional linear search. While the inner-layer optimization problem is transformed into a convex SDP problem with SDR technique and Charnes-Cooper transformation. In the perfect CSI case of both non-colluding and colluding Eves scenarios, we prove that the relaxation of SDR is tight and analyze the complexity of proposed algorithms. Finally, simulation results validate the effectiveness and robustness of proposed scheme.

• Applied Microscopy
• /
• v.14 no.2
• /
• pp.38-51
• /
• 1984

Fine structure of dormant and swollen conidiospore from Trichoderma koningii and the mechanism of protoplasting from the conidiospore were studied by scanning and transmission electron microscopy. The cell wall of dormant conidiospore was two-layered structure which consisted of electron dense outer layer and electron transparent inner layer. After 8.5 hrs incubation. the conidiospore was swollen and the outer layer of cell wall shown unequal thickness and partial breakage. Protoplast was released through the pore which has been formed by the breakage of outer layer and dissolution of newly synthesized cell wall for germ-tube formation. Swollen conidiospore and protoplast in releasing process contained various cell organelles and vacuoles with electron dense materials. The protoplast contained looser cytoplasm and had no cell wall materials outside of plasmamembrane.

• The Journal of Engineering Geology
• /
• v.33 no.2
• /
• pp.293-305
• /
• 2023

Three ground models are analyzed using a 1g shaking table and laminar shear box (LSB) to investigate the impact of the ground structure on seismic wave amplification during earthquakes. Multi-layer horizontal, embankment, and basin ground models are selected for this investigation, with each model being divided into dense and loose ground layers, Accelerometers are installed during the construction of each ground model to capture any seismic wave amplification owing th the propagation of an artificial seismic wave, sine wave sweep, and 10-Hz sine wave through a given ground model. The amplification of the tested seismic waves is analyzed using the observed peak ground acceleration and spectrum acceleration. The observed acceleration amplification in the multi-layer horizontal ground model is significantly higher the seismic waves that propagated across the dense ground-loose ground boundary compared with those that only propagated through the dense ground. Furthermore, the observed acceleration amplification gradually increases in the central part of the multi-layer embankment and basin models for the seismic waves that propagated across the dense ground-loose ground boundary.

• Applied Microscopy
• /
• v.28 no.3
• /
• pp.273-282
• /
• 1998

Ultrastructure of fertilized egg envelope in the tomato clown anemonefish (Amphiprion freatus) was examined by mean of light, scanning and transmission electron microscopies. The fertilized egg of tomato clown anemonefish was of non-transparent, ellipsoidal, adhesive and non-floted type, and there were numerous oil droplets in the yolk sac and adhesive filaments in the area of the animal pole. The outer surface of egg envelope was smooth. The fertilized egg envelope consisted of two distinct layers; a non-adhesive outer electron-dense layer and an inner layer, consisting of five horizontal low electron-dense lamellae alternating with the middle electron dense interlamellae.

• 한국정보디스플레이학회:학술대회논문집
• /
• 2005.07b
• /
• pp.1499-1502
• /
• 2005

Dense $BaMgAl_{10}O_{17}:Eu^{2+}$ phosphor particles with a spherical shape have been synthesized through spray pyrolysis method using basic aluminum nitrate precursor as a spray solution. This $BaMgAl_{10}O_{17}:Eu^{2+}$ particles prepared by the spray pyrolysis have shown the stronger emission intensity compared to the commercially-available $BaMgAl_{10}O_{17}:Eu^{2+}$. However, thermal stability of the BAM:Eu b lue phosphor is very poor due to changing from $Eu^{2+}$ to $Eu^{3+}$ at the thermal process, so brightness of the phosphor decreases. To improve the thermal stability of the dense BAM:Eu phosphor, the spherical BAM:Eu particles were coated with pure $BaMgAl_{10}O_{17}$ layer using the hydrolysis reaction in a solution system. The synthesized powders were characterized by XRD, SEM and PL. On the other hand, the emission properties of the BAM:Eu phosphors coated with $BaMgAl_{10}O_{17}$ layer before and after thermal treatment at $500^{\circ}C$ for 30 min were estimated under VUV excitation. The brightness of the coated phosphor was higher than that of the uncoated phosphor. Also, the coating thickness of BAM layer in the BAM:Eu particles was optimized.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.13 no.2
• /
• pp.494-513
• /
• 2019

A multi-channel access problem based on hypergraph model in ultra-dense small cell networks is studied in this paper. Due to the hyper-dense deployment of samll cells and the low-powered equipment, cumulative interference becomes an important problem besides the direct interference. The traditional binary interference model cannot capture the complicated interference relationship. In order to overcome this shortcoming, we use the hypergraph model to describe the cumulative interference relation among small cells. We formulate the multi-channel access problem based on hypergraph as two local altruistic games. The first game aims at minimizing the protocol MAC layer interference, which requires less information exchange and can converge faster. The second game aims at minimizing the physical layer interference. It needs more information interaction and converges slower, obtaining better performance. The two modeled games are both proved to be exact potential games, which admit at least one pure Nash Equilibrium (NE). To provide information exchange and reduce convergecne time, a cloud-based centralized-distributed algorithm is designed. Simulation results show that the proposed hypergraph models are both superior to the existing binary models and show the pros and cons of the two methods in different aspects.

• Journal of Korean Society for Atmospheric Environment
• /
• v.21 no.5
• /
• pp.537-545
• /
• 2005

A new dispersion model for dense gas is constructed in the Lagrangian framework. Prediction of concentration by the proposed model is compared with measure data obtained in the experiment conducted in Thorney Island in 1984. Two major effects of dense gas dispersion, gravity slumping and stratification effect, are successfully incorporated into LDM (Lagrangian dense gas model). Entrainment effect is naturally modelled by introducing stochastic dispersion model with the effect of turbulence suppression by stratification. Not only various releasing conditions but also complex terrain can be extended to, although proposed model is appropriate for flat terrain.

• Journal of Korea Water Resources Association
• /
• v.50 no.9
• /
• pp.637-646
• /
• 2017

This study presents a numerical model for simulating dense interflows. The governing equations are provided and the finite difference method is used with the $k-{\varepsilon}$ turbulence model. The model is used to simulate a dense interflow established in a deep ambient water, resulting velocity and excess density profiles. It is observed that velocity decreases in the longitudinal direction due to water entrainment in the vicinity of the outlet and rarely changes for increased Richardson number. Similarity collapses of velocity and excess density are obtained, but those of turbulent kinetic energy and dissipation rate are not. A shape factor for the dense interflow is obtained from the simulated profiles. The value of this shape factor can be used in the layer-averaged modeling of dense interflows. In addition, a buoyancy-related parameter ($c_{3{\varepsilon}}$) for the $k-{\varepsilon}$ model and the volume expansion coefficient (${\beta}_0$) are obtained from the simulated results.

• The Journal of Korean Institute of Information Technology
• /
• v.17 no.8
• /
• pp.11-19
• /
• 2019

Against advanced image-related crimes, a high level of digital forensic methods is required. However, feature-based methods are difficult to respond to new device features by utilizing human-designed features, and deep learning-based methods should improve accuracy. This paper proposes a deep learning model to identify camera models based on DenseNet, the recent technology in the deep learning model field. To extract camera sensor features, a HPF feature extraction filter was applied. For camera model identification, we modified the number of hierarchical iterations and eliminated the Bottleneck layer and compression processing used to reduce computation. The proposed model was analyzed using the Dresden database and achieved an accuracy of 99.65% for 14 camera models. We achieved higher accuracy than previous studies and overcome their disadvantages with low accuracy for the same manufacturer.