• Journal of the Korea Society of Computer and Information
• /
• v.25 no.6
• /
• pp.35-47
• /
• 2020

Box-office prediction is important to movie stakeholders. It is necessary to accurately predict box-office and select important variables. In this paper, we propose a multivariate time series classification and important variable selection method to improve accuracy of predicting the box-office. As a research method, we collected daily data from KOBIS and NAVER for South Korean movies, selected important variables using Random Forest and predicted multivariate time series using Deep Learning. Based on the Korean screen quota system, Deep Learning was used to compare the accuracy of box-office predictions on the 73^rd day from movie release with the important variables and entire variables, and the results was tested whether they are statistically significant. As a Deep Learning model, Multi-Layer Perceptron, Fully Convolutional Neural Networks, and Residual Network were used. Among the Deep Learning models, the model using important variables and Residual Network had the highest prediction accuracy at 93%.

• Structural Engineering and Mechanics
• /
• v.32 no.2
• /
• pp.193-212
• /
• 2009

The resilience of a city confronted with a terrorist bomb attack is the background of the paper. The resilience strongly depends on vital infrastructure and the physical protection of people. The protection buildings provide in case of an external explosion is one of the important elements in safety assessment. Besides the aspect of protection, buildings facilitate and enable many functions, e.g., offices, data storage, -handling and -transfer, energy supply, banks, shopping malls etc. When a building is damaged, the loss of functions is directly related to the location, amount of damage and the damage level. At TNO Defence, Security and Safety methods are developed to quantify the resilience of city infrastructure systems (Weerheijm et al. 2007b). In this framework, the dynamic response, damage levels and residual bearing capacity of multi-storey RC buildings is studied. The current paper addresses the aspects of dynamic response and progressive collapse, as well as the proposed method to relate the structural damage to a volume-damage parameter, which can be linked to the loss of functionality. After a general introduction to the research programme and progressive collapse, the study of the dynamic response and damage due to blast loading for a single RC element is described. Shock tube experiments on plates are used as a reference to study the possibilities of engineering methods and an explicit finite element code to quantify the response and residual bearing capacity. Next the dynamic response and progressive collapse of a multi storey RC building is studied numerically, using a number of models. Conclusions are drawn on the ability to predict initial blast damage and progressive collapse. Finally the link between the structural damage of a building and its loss of functionality is described, which is essential input for the envisaged method to quantify the resilience of city infrastructure.

• Transactions of the Korean Society of Mechanical Engineers A
• /
• v.33 no.4
• /
• pp.310-320
• /
• 2009

In this paper, we estimate area-averaged solution of peening residual stress using a 3-D multi-impact symmetry-cell FE model. The symmetry-cell model includes factors reflecting peening phenomena and plastic shot. Area-averaged solution is much closer to XRD experimental solution than 4-node-averaged solution in plastic shot FE model. We then obtain FE Almen saturation curve corresponding to experimental Almen curve based on area-averaged solution. Using the curve, we obtain FE area-averaged solution in major peening materials, and compare the FE solution with experimental solution. In peening materials, surface, maximum compressive residual stress and deformation depth reach experimental solutions. Thus, FE Almen curve is useful for estimation of residual stress solution and could improve the efficiency of peening process. Consequently, it is confirmed that concept of area-averaged solution is the realistic analytical method for evaluation of peening residual stress.

• 한국HCI학회:학술대회논문집
• /
• 2008.02a
• /
• pp.190-194
• /
• 2008

This paper presents a speaker identification technique which is one of the basic techniques of the ubiquitous robot companion. Though the conventional mel-frequency cepstral coefficients guarantee high performance of speaker identification in clean condition, the performance is degraded dramatically in noise condition. To overcome this problem, we employed the relative autocorrelation sequence mel-frequency cepstral coefficient which is one of the noise robust features. However, there are two problems in relative autocorrelation sequence mel-frequency cepstral coefficient: 1) the limited information problem. 2) the residual noise problem. In this paper, to deal with these drawbacks, we propose a multi-streaming method for the limited information problem and a hybrid method for the residual noise problem. To evaluate proposed methods, noisy speech is used in which air conditioner noise, classic music, and vacuum noise are artificially added. Through experiments, proposed methods provide better performance of speaker identification than the conventional methods.

• Proceedings of the KIEE Conference
• /
• 2005.10c
• /
• pp.50-52
• /
• 2005

In this paper, the analysis of the eddy current testing(ECT) signals under thc Influence of the ferromagnetic support plate was performed in steam generator(SG) tube of nuclear power plant. In order to remove the influence of the ferromagnetic support plate, a multi-frequency ECT was used. The models which was established for the analysis of the signals is calculated using numerical analysis of finite element method. Through the result of numerical analysis, improved signals is acquired considering the influence of the ferromagnetic support plate using mixing of multi-frequency This paper is presented the residual errors and the phase changes for analysis of the defect signals which should be considered when conducting a ECT using multi-frequency.

• International Journal of Aeronautical and Space Sciences
• /
• v.18 no.4
• /
• pp.601-613
• /
• 2017

An adjoint-based error estimation and mesh adaptation study is conducted for two-dimensional viscous flows on unstructured hybrid meshes. The error in an integral output functional of interest is estimated by a dot product of the residual vector and adjoint variable vector. Regions for the mesh to be adapted are selected based on the amount of local error at each nodal point. Triangular cells in the adaptive regions are refined by regular refinement, and quadrangular cells near viscous walls are bisected accordingly. The present procedure is applied to single-element airfoils such as the RAE2822 at a transonic regime and a diamond-shaped airfoil at a supersonic regime. Then the 30P30N multi-element airfoil at a low subsonic regime with a high incidence angle (${\alpha}=21deg.$) is analyzed. The same level of prediction accuracy for lift and drag is achieved with much less mesh points than the uniform mesh refinement approach. The detailed procedure of the adjoint-based mesh refinement for the multi-element airfoil case show that the basic flow features around the airfoil should be resolved so that the adjoint method can accurately estimate an output error.

• Proceedings of the Korean Society for Technology of Plasticity Conference
• /
• 2008.05a
• /
• pp.220-223
• /
• 2008

Large curved plate blocks are widely used to construct hull structure in shipbuilding industry. Most curved plates are manufactured by using manual method called as line heating that use deformation caused by residual stress after local heating along a line which is perpendicular to the curvature direction. However, its working environment is poor and its formability is totally dependent on an experienced technician. In view of that, multi-point dieless forming (MDF) technology that use reconfigurable punch arrays instead of one piece die is proposed in this study. The MDF process is based on a concept of equivalent die surface made by numbers of punches which has round tip at the end of it. In this study, numerical simulation for common curvature type such as saddle shape was carried out. In addition, experiments in the plate forming process were also conducted to compare with the numerical results in view of final configuration. Consequently, it was noted that the proposed dieless forming method has considerable feasibility to substitute the new process for conventional manual method.

• 한국전산유체공학회:학술대회논문집
• /
• 2004.03a
• /
• pp.83-90
• /
• 2004

Substructuring methods are usually used in finite element structural analyses. In this study a multi-level substructuring algorithm is developed and proposed as a possible candidate for incompressible fluid solves. Finite element formulation for incompressible flow has been stabilized by a modified residual procedure proposed by Ilinca et.al.[5]. The present algorithm consists of four stages such as a gathering stage, a condensing stage, a solving stage and a scattering stage. At each level, a predetermined number of elements are gathered and condensed to form an element of higher level. At highest level, each subdomain consists of only one super-element. Thus, the inversion process of a stiffness matrix associated with internal degrees of freedom of each subdomain has been replaced by a sequential static condensation. The global algebraic system arising feom the assembly of each subdomains is solved using Conjugate Gradient Squared(CGS) method. In this case, pre-conditioning techniques usually accompanied by iterative solvers are not needed.

• Proceedings of the KSRS Conference
• /
• 2003.11a
• /
• pp.888-890
• /
• 2003

The objective of this investigation is to establish a simple yet effective block adjustment procedure for the orthorectification of multi-orbit satellite images. The major works of the proposed scheme are: (1) adjustment of satellite‘s orbit accurately, (2) calculation of the error vectors for each tie point using digital terrain model and ray tracing technique, (3) refining the orbit using the Least Squares Filtering technique and (4) generation of the orthophotos. In the process of least squares filtering, we use the residual vectors on ground control points and tie points to collocate the orbit. In orthorectification, we use the indirect method to generate the orthoimage. Test areas cover northern Taiwan. Test images are from SPOT 5 satellite. Experimental results indicate that proposed method improves the relative accuracy significantly.

• Journal of Korea Multimedia Society
• /
• v.25 no.11
• /
• pp.1582-1590
• /
• 2022

Due to the situation of the widespread of the coronavirus, which causes the problem of lack of face image data occluded by masks at recent time, in order to solve the related problems, this paper proposes a method to generate face images with masks using a combination of generative adversarial networks and spatial transformation networks based on CNN model. The system we proposed in this paper is based on the GAN, combined with multi-scale convolution kernels to extract features at different details of the human face images, and used Wasserstein divergence as the measure of the distance between real samples and synthetic samples in order to optimize Generator performance. Experiments show that the proposed method can effectively put masks on face images with high efficiency and fast reaction time and the synthesized human face images are pretty natural and real.