• ETRI Journal
• /
• v.39 no.3
• /
• pp.398-405
• /
• 2017

In this paper, an artificial stereo extension method that creates stereophonic sound from a mono sound source is proposed. The proposed method first trains deep neural networks (DNNs) that model the nonlinear relationship between the dominant and residual signals of the stereo channel. In the training stage, the band-wise log spectral magnitude and unwrapped phase of both the dominant and residual signals are utilized to model the nonlinearities of each sub-band through deep architecture. From that point, stereo extension is conducted by estimating the residual signal that corresponds to the input mono channel signal with the trained DNN model in a sub-band domain. The performance of the proposed method was evaluated using a log spectral distortion (LSD) measure and multiple stimuli with a hidden reference and anchor (MUSHRA) test. The results showed that the proposed method provided a lower LSD and higher MUSHRA score than conventional methods that use hidden Markov models and DNN with full-band processing.

• Wind and Structures
• /
• v.24 no.3
• /
• pp.223-247
• /
• 2017

By using numerical simulation, vast and detailed information and observation of the physics of flow over a train model can be obtained. However, the accuracy of the numerical results is questionable as it is affected by grid convergence error. This paper describes a systematic method of computational grid refinement for the Unsteady Reynolds Navier-Stokes (URANS) of flow around a generic model of trains using the OpenFOAM software. The sensitivity of the computed flow field on different mesh resolutions is investigated in this paper. This involves solutions on three different grid refinements, namely fine, medium, and coarse grids to investigate the effect of grid dependency. The level of grid independence is evaluated using a form of Richardson extrapolation and Grid Convergence Index (GCI). This is done by comparing the GCI results of various parameters between different levels of mesh resolutions. In this study, monotonic convergence criteria were achieved, indicating that the grid convergence error was progressively reduced. The fine grid resolution's GCI value was less than 1%. The results from a simulation of the finest grid resolution, which includes pressure coefficient, drag coefficient and flow visualization, are presented and compared to previous available data.

• Proceedings of the KSR Conference
• /
• 2003.05a
• /
• pp.337-342
• /
• 2003

The dynamic load effects are conveyed to the railway bridges through tracks which are generated by moving trains The dynamic load effects may vary due to the dynamic characteristics of the applied vehicle loads and the railway bridges containing the track system. However, the track effects have been neglected or simplified by spring elements in the most studies since it is quite complex to consider the track systems in the dynamic analysis models of railway bridges. In this study, track system on railway bridges is modeled using a three-dimensional discrete-support model that can simulate the load carrying behavior of tracks. In addition, this program is developed with the precise 20-car model and a continuous PSC(prestressed concrete) box girder bridge, which is the main bridge type of Korea Train express(KTX). Three-dimensional elements are used for both. The dynamic response of railway bridges is found to be affected depending on whether the track model is considered or not. The influencing rate depends on the traveling speed and different wheel-axle distance. The dynamic bridge response is decreased remarkably by the track systems around the resonant frequency. Therefore, the resonance effect can be reduced by modifying the track properties in the railway bridge.

• Wind and Structures
• /
• v.36 no.1
• /
• pp.1-14
• /
• 2023

To evaluate the wind shielding effect of bridge towers with multiple limbs on high-speed trains, a wind tunnel test was conducted to investigate the aerodynamic characteristics of vehicles traversing multi-limb towers, which represented a combination of the steady aerodynamic coefficient of the vehicle-bridge system and wind environment around the tower. Subsequently, the analysis model of wind-vehicle-bridge (WVB) system considering the additional moments caused by lift and drag forces under nonuniform wind was proposed, and the reliability and accuracy of the proposed model of WVB system were verified using another model. Finally, the factors influencing the wind shielding effect of multi-limb towers were analyzed. The results indicate that the wind speed distributions along the span exhibit two sudden changes, and the wind speed generally decreases with increasing wind direction angle. The pitching and yawing accelerations of vehicles under nonuniform wind loads significantly increase due to the additional pitching and yawing moments. The sudden change values of the lateral and yawing accelerations caused by the wind shielding effect of multi-limb tower are 0.43 m/s² and 0.11 rad/s² within 0.4 s, respectively. The results indicate that the wind shielding effect of a multi-limb tower is the controlling factor in WVB systems.

• IEMEK Journal of Embedded Systems and Applications
• /
• v.19 no.2
• /
• pp.91-99
• /
• 2024

Collision avoidance system is important to improve the robustness and functional safety of autonomous vehicles. This paper proposes an object-level distance estimation method to develop a collision avoidance system, and it is applied to golfcarts utilized in country club environments. To improve the detection accuracy, we continually trained an object detection model based on pseudo labels generated by a pre-trained detector. Moreover, we propose object-aware depth estimation (OADE) method which trains a depth model focusing on object regions. In the OADE algorithm, we generated dense depth information for object regions by utilizing detection results and sparse LiDAR points, and it is referred to as object-aware LiDAR projection (OALP). By using the OALP maps, a depth estimation model was trained by backpropagating more gradients of the loss on object regions. Experiments were conducted on our custom dataset, which was collected for the travel distance of 22 km on 54 holes in three country clubs under various weather conditions. The precision and recall rate were respectively improved from 70.5% and 49.1% to 95.3% and 92.1% after the continual learning with pseudo labels. Moreover, the OADE algorithm reduces the absolute relative error from 4.76% to 4.27% for estimating distances to obstacles.

• Journal of the Computational Structural Engineering Institute of Korea
• /
• v.15 no.4
• /
• pp.693-702
• /
• 2002

The aim of this study is to develop a 3-dimensional dynamic analysis model, capable of considering the interaction between vehicles and bridges more accurately. The dynamic analysis model is developed with the high-speed train (KTX) and a 2-span continuous prestressed concrete box girder bridge with a double track. The 20-car model is developed using the moving vehicle model for the regular trainset. Three-dimensional frame elements are used for the bridge model. Using the developed models, a dynamic behavior analysis program is coded. The analytical results are compared with the dynamic field test results and found to be valid to yield quite accurate dynamic responses. Based on the results of this study, the hybrid model, made up of the moving vehicle model for the heaviest power car and the moving force model for the other cars, is quite simple and effective without loosing the accuracy that much. Under the coincidence condition of two trains traveling with resonance velocity in the opposite directions, it is necessary to check not only the dynamic responses of the bridge with one-way traffic but those with two- way coincidence.

• Journal of the Korean Society for Railway
• /
• v.10 no.6
• /
• pp.678-684
• /
• 2007

This research has defined the factors of noise in cars during subway train services, which is surfacing as a new environmental trouble. It shows additional accomplishment of a discerning analysis on the standard of noise regulation as well as its seriousness. According to the Enforcement Regulations for Noise and Vibration under the Ministry of Environment and its standard noise regulation figure 70dB, we divided two groups of which train noise figures are over and under 70dB respectively, and used their 359 results about noise, geometric structures and operation elements, for this analysis. The results and suggestions are following. First of all, when we discern the seriousness of noise in a train, the track type has mattered in geometric structure and the velocity in operation elements. Therefore, when we construct subway from now on, we should take the track type in consideration and establish plans to keep proper speed in respect of operation. Secondly, the established discernment model in this research can be used in making alternative plans or improvement of subway trains hereafter, showing relatively high accuracy of estimation. Consequently, the readjustment of geometric structure and operation elements is needed, not to make it over the regulation standard of noise in case the noise in train is serious. The discriminant model of this research can be used as elementary material for comfortable and safe subway trains, making the estimation of noise seriousness possible.

• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.22 no.5
• /
• pp.7-14
• /
• 2021

Tomato crops are highly affected by tomato diseases, and if not prevented, a disease can cause severe losses for the agricultural economy. Therefore, there is a need for a system that quickly and accurately diagnoses various tomato diseases. In this paper, we propose a system that classifies nine diseases as well as healthy tomato plants by applying various pretrained deep learning-based CNN models trained on an ImageNet dataset. The tomato leaf image dataset obtained from PlantVillage is provided as input to ResNet, Xception, and DenseNet, which have deep learning-based CNN architectures. The proposed models were constructed by adding a top-level classifier to the basic CNN model, and they were trained by applying a 5-fold cross-validation strategy. All three of the proposed models were trained in two stages: transfer learning (which freezes the layers of the basic CNN model and then trains only the top-level classifiers), and fine-tuned learning (which sets the learning rate to a very small number and trains after unfreezing basic CNN layers). SGD, RMSprop, and Adam were applied as optimization algorithms. The experimental results show that the DenseNet CNN model to which the RMSprop algorithm was applied output the best results, with 98.63% accuracy.

• Wind and Structures
• /
• v.20 no.2
• /
• pp.143-168
• /
• 2015

A numerical model for analyzing air-train-track interaction is proposed to investigate the dynamic behavior of a high-speed train running on a track in crosswinds. The model is composed of a train-track interaction model and a train-air interaction model. The train-track interaction model is built on the basis of the vehicle-track coupled dynamics theory. The train-air interaction model is developed based on the train aerodynamics, in which the Arbitrary Lagrangian-Eulerian (ALE) method is employed to deal with the dynamic boundary between the train and the air. Based on the air-train-track model, characteristics of flow structure around a high-speed train are described and the dynamic behavior of the high-speed train running on track in crosswinds is investigated. Results show that the dynamic indices of the head car are larger than those of other cars in crosswinds. From the viewpoint of dynamic safety evaluation, the running safety of the train in crosswinds is basically controlled by the head car. Compared with the generally used assessment indices of running safety such as the derailment coefficient and the wheel-load reduction ratio, the overturning coefficient will overestimate the running safety of a train on a track under crosswind condition. It is suggested to use the wheel-load reduction ratio and the lateral wheel-rail force as the dominant safety assessment indices when high-speed trains run in crosswinds.

• KNOM Review
• /
• v.24 no.2
• /
• pp.24-34
• /
• 2021

Service providers using edge computing provide a high level of service. As a result, devices store important information in inner storage and have become a target of the latest cyberattacks, which are more difficult to detect. Although experts use a security system such as intrusion detection systems, the existing intrusion systems have low detection accuracy. Therefore, in this paper, we proposed a machine learning model for more accurate intrusion detections of devices in edge computing. The proposed model is a hybrid model that combines a stacked sparse autoencoder (SSAE) and a convolutional neural network (CNN) to extract important feature vectors from the input data using sparsity constraints. To find the optimal model, we compared and analyzed the performance as adjusting the sparsity coefficient of SSAE. As a result, the model showed the highest accuracy as a 96.9% using the sparsity constraints. Therefore, the model showed the highest performance when model trains only important features.