• Proceedings of the Korea Technology Innovation Society Conference
• /
• 2013.10a
• /
• pp.88-97
• /
• 2013

• Journal of IKEEE
• /
• v.23 no.4
• /
• pp.1295-1301
• /
• 2019

In this paper, we propose a novel method of denoising human motion using a bidirectional recurrent neural network (BRNN) with an attention mechanism. The corrupted motion captured from a single 3D depth sensor camera is automatically fixed in the well-established smooth motion manifold. Incorporating an attention mechanism into BRNN achieves better optimization results and higher accuracy than other deep learning frameworks because a higher weight value is selectively given to a more important input pose at a specific frame for encoding the input motion. Experimental results show that our approach effectively handles various types of motion and noise, and we believe that our method can sufficiently be used in motion capture applications as a post-processing step after capturing human motion.

• The Bulletin of The Korean Astronomical Society
• /
• v.37 no.1
• /
• pp.85.2-85.2
• /
• 2012

In this paper, we introduce the 3D modeling of the coronal magnetic field in the solar active region by extrapolating from the 2D observational data numerically. First, we introduce a nonlinear force-free field (NLFFF) extrapolation code based on the MHD-like relaxation method implementing the cleaning a numerical error for Div B proposed by Dedner et al. 2002 and the multi-grid method. We are able to reconstruct the ideal force-free field, which was introduced by Low & Lou (1990), in high accuracy and achieve the faster speed in the high-resolution calculation (512^3 grids). Next we applied our NLFFF extrapolation to the solar active region NOAA 10930. First of all, we compare the 3D NLFFF with the flare ribbons of Ca II images observed by the Solar Optical Telescope (SOT) aboard on the Hinode. As a result, it was found that the location of the two foot-points of the magnetic field lines well correspond to the flare ribbon. The result indicates that the NLFFF well capture the 3D structure of magnetic field in the flaring region. We further report the stability of the magnetic field by estimating the twist value of the field line and finally suggest the flare onset mechanism.

• International Journal of Aeronautical and Space Sciences
• /
• v.16 no.4
• /
• pp.560-570
• /
• 2015

We present a simulation of a hybrid Reynolds-averaged Navier Stokes / Large Eddy Simulation (RANS/LES) based on detached eddy simulation (DES) for a Burrows and Kurkov supersonic planar mixing experiment. The preliminary simulation results are checked in order to validate the numerical computing capability of the current code. Mesh refinement studies are performed to identify the minimum grid size required to accurately capture the flow physics. A detailed investigation of the turbulence/chemistry interaction is carried out for a nine species 19-step hydrogen-air reaction mechanism. In contrast to the instantaneous value, the simulated time-averaged result inside the reactive shear layer underpredicts the maximum rise in $H_2O$ concentration and total temperature relative to the experimental data. The reason for the discrepancy is described in detail. Combustion parameters such as OH mass fraction, flame index, scalar dissipation rate, and mixture fraction are analyzed in order to study the flame structure.