• Journal of information and communication convergence engineering
• /
• v.21 no.3
• /
• pp.208-215
• /
• 2023

Deep learning techniques provide powerful solutions to several pattern-recognition problems, including Raman spectral classification. However, these networks require large amounts of labeled data to perform well. Labeled data, which are typically obtained in a laboratory, can potentially be alleviated by data augmentation. This study investigated various data augmentation techniques and applied multiple deep learning methods to Raman spectral classification. Raman spectra yield fingerprint-like information about chemical compositions, but are prone to noise when the particles of the material are small. Five augmentation models were investigated to build robust deep learning classifiers: weighted sums of spectral signals, imitated chemical backgrounds, extended multiplicative signal augmentation, and generated Gaussian and Poisson-distributed noise. We compared the performance of nine state-of-the-art convolutional neural networks with all the augmentation techniques. The LeNet5 models with background noise augmentation yielded the highest accuracy when tested on real-world Raman spectral classification at 88.33% accuracy. A class activation map of the model was generated to provide a qualitative observation of the results.

• Atmosphere
• /
• v.33 no.4
• /
• pp.387-398
• /
• 2023

The double Fourier series (DFS) spectral dynamical core is evaluated for the two idealized test cases in comparison with the spherical harmonics (SPH) spectral dynamical core. A new approach in calculating the meridional expansion coefficients of DFS, which was recently developed to alleviate a computational error but only applied to the 2D spherical shallow water equation, is also tested. In the 3D deformational tracer transport test, the difference is not conspicuous between SPH and DFS simulations, with a slight outperformance of the new DFS approach in terms of undershooting problem. In the baroclinic wave development test, the DFS-simulated wave pattern is quantitatively similar to the SPH-simulated one at high resolutions, but with a substantially lower computational cost. The new DFS approach does not offer a salient advantage compared to the original DFS while computation cost slightly increases. This result suggests that the current DFS spectral method can be a practical and alternative dynamical core for high-resolution global modeling.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2020.11a
• /
• pp.87-90
• /
• 2020

GPU의 발전과 함께 성장한 딥러닝(Deep Learning)은 영상 분류 문제에서 최고의 성능을 보이고 있다. 그러나 합성곱 신경망 기반의 모델을 깊게 쌓음에 따라 신경망의 표현력이 좋아짐과 동시에 때로는 학습이 잘되지 않고 성능이 저하되는 등의 부작용도 등장했다. 성능 향상을 방해하는 주요 요인 중 하나는, 차원감소 목적에 따라 필연적으로 정보 손실을 겪어야 하는 풀링 계층에 있다. 따라서 특성맵(Feature map)의 차원감소를 통해 얻게 되는 비용적 이득과 모델의 분류 성능 사이의 취사선택(Trade-off)이 존재한다. 그리고 이로부터 자유로워지기 위한 다양한 연구와 기법이 존재하는데 Spectral Pooling도 이 중 하나이다. 본 논문에서는 이산 푸리에 변환(Discrete Fourier Transform, DFT)을 이용한 Spectral Pooling에 대한 소개와, 해당 풀링의 성질을 통상적으로 사용되고 있는 Max Pooling과의 성능 비교를 통해 분석한다. 또한 영상 내 고주파수 부분에서 특히 더 강건하지 못하다는 맥스 풀링의 고질적인 문제점을, Spectral Pooling과의 하이브리드(Hybrid) 구조를 통해 어떻게 극복해나갈 것인지 그 가능성을 중심으로 실험을 수행했다.

• Earthquakes and Structures
• /
• v.25 no.5
• /
• pp.325-342
• /
• 2023

The damping modification factor (DMF) is used to modify the 5%-damped response spectrum to produce spectral values that correspond to other necessary damping ratios for seismic design. The DMF has been the subject of numerous studies, and it has been discovered that seismological parameters like magnitude and distance can have an impact on it. However, DMF formulations incorporating these seismological parameters cannot be directly applied to seismic design because these parameters are not specified in the present seismic codes. The goal of this study is to develop a formulation for the DMF that can be directly applied in seismic design and that takes the effects of magnitude, distance, and site conditions into account. To achieve this goal, 16660 ground motions with magnitudes ranging from 4 to 9 and epicentral distances ranging from 10 to 200 km are used to systematically study the effects of magnitude, distance, and site conditions on the DMF. Furthermore, according to the knowledge that magnitude and distance affect the DMF primarily by changing the spectral shape, a spectral shape factor is adopted to reflect influences of magnitude and distance, and a new formulation for the DMF incorporating the spectral shape factor is developed. In comparison to the current formulations, the proposed formulation provides a more accurate prediction of the DMF and can be employed directly in seismic design.

• The Journal of the Acoustical Society of Korea
• /
• v.28 no.3
• /
• pp.273-278
• /
• 2009

In this paper, we propose the spectral transition rate (STR) as a novel feature for speech and music discrimination (SMD). We observed that the spectral peaks of speech signal are gradually changing due to coarticulation effect. However, the sound of musical instruments in general tends to keep the peak frequencies and energies unchanged for relatively long period of time compared to speech. The STR of speech is much higher than that of music. The experimental results show that the STR based SMD method outperforms a conventional method. Especially, the STR based SMD gives relatively fast output without any performance degradation.

• Journal of the Chungcheong Mathematical Society
• /
• v.16 no.1
• /
• pp.37-48
• /
• 2003

In this paper, we show that for a pure hyponormal operator the analytic spectral subspace and the algebraic spectral subspace are coincide. Using this result, we have the following result: Let T be a decomposable operator on a Banach space X and let S be a pure hyponormal operator on a Hilbert space H. Then every linear operator ${\theta}:X{\rightarrow}H$ with $S{\theta}={\theta}T$ is automatically continuous.

• Journal of the Korean Mathematical Society
• /
• v.38 no.3
• /
• pp.633-644
• /
• 2001

Revenel computed the Adams spectral sequence converging to BP(Ω$^2$S(sup)2n+1) and got the E(sub)$\infty$-term. Then he gave the conjecture about the extension. Here we prove that there should be non-trivial extension. We also study the BP(sub)*BP comodule structures on the polynomial algebras which are related with BP(sub)*(Ω$^2$S(sup)2n+1).

• Proceedings of the Computational Structural Engineering Institute Conference
• /
• 2008.04a
• /
• pp.45-48
• /
• 2008

Periodic lattice structures such as the large space lattice structures and carbon nanotubes may take the extension-transverse shear-bending coupled vibrations, which can be well represented by the extended Timoshenko beam theory. In this paper, the spectrally formulated finite element model (simply, spectral element model) has been developed for extended Timoshenko beams and applied to some typical periodic lattice structures such as the armchair carbon nanotube, the periodic plane truss, and the periodic space lattice beam.

• Journal of the Korean Mathematical Society
• /
• v.35 no.3
• /
• pp.637-658
• /
• 1998

One-dimensional diffusion processes are characterized by Feller's data of canonical scales and speed measures and, if we apply the theory of spectral functions of strings developed by M. G. Krein, Feller's data are determined by paris of spectral characteristic functions so that theses pairs may be considered as invariants of diffusions under the homeomorphic change of state spaces. We show by examples how these invariants are useful in the study of one-dimensional diffusion processes.

• Proceedings of the Optical Society of Korea Conference
• /
• 2009.02a
• /
• pp.169-170
• /
• 2009

We present a quantification protocol to get the alignment factors of a custom-made spectrometer and the nonlinear fitting function between the measured CCD pixel domain and the wavelength domain to apply the spectral-domain optical coherence tomography using fiber Bragg gratings. We have used 4 different center wavelength gratings with a narrow spectral bandwidth (<0.05 nm) and the same reflectivity (>92 %) to calibrate and align the custom-made spectrometer.