• 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.

• 한국지구물리탐사학회:학술대회논문집
• /
• 2008.10a
• /
• pp.107-112
• /
• 2008

In the case of the deep reservoirs like the gas reservoirs in the East-sea, it is often difficult to observe AVO responses in CMP gathers. Because the reservoir becomes more consolidated as its depth deepens, P-wave velocity does not decrease significantly when the pore fluid is replaced by the gas. In this study, we analyzed the effects of Poisson's ratio difference on AVO response with a variety of Poisson's ratios for the upper and lower layers. The results show that, as the difference in Poisson's ratio between the upper and lower layers decreases, the change in the reflection amplitude with incidence angle decreases. To consider the limitation of AVO responses shown in the gas reservoir in East-sea, the velocity model was made by simulation Gorae V structure with seismic data and well logs. The results of comparing AVO responses observed from the synthetic data with theoretical AVO responses calculated by using material properties show that the amount of the change in reflection amplitude with increasing incident angle is very small when the difference in Poisson's ratio between the upper and lower layers is small. In addition, the characteristics of AVO responses were concealed by noise or amplitude distortion arisen during preprocessing. To overcome such limitations of AVO analysis of the data from deep reservoirs, we need to acquire precisely reflection amplitudes in data acquisition stage and use processing tools which preserve reflection amplitude in data processing stage.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.11 no.2
• /
• pp.628-649
• /
• 2017

Security and resource-saving are both demands of the fifth generation (5G) wireless networks. In this paper, we study the secrecy spectrum efficiency (SSE) and secrecy energy efficiency (SEE) of a K-tier massive multiple-input multiple-output (MIMO) enabled heterogeneous cellular network (HetNet), in which artificial noise (AN) are employed for secrecy enhancement. Assuming (i) independent Poisson point process model for the locations of base stations (BSs) of each tier as well as that of eavesdroppers, (ii) zero-forcing precoding at the macrocell BSs (MBSs), and (iii) maximum average received power-based cell selection, the tractable lower bound expressions for SSE and SEE of massive MIMO enabled HetNets are derived. Then, the influences on secrecy oriented spectrum and energy efficiency performance caused by the power allocation for AN, transmit antenna number, number of users served by each MBS, and eavesdropper density are analyzed respectively. Moreover, the analysis accuracy is verified by Monte Carlo simulations.

• Journal of the Korea Society of Computer and Information
• /
• v.28 no.3
• /
• pp.25-33
• /
• 2023

Noise generated during the acquisition and transmission of CT images acts as a factor that degrades image quality. Therefore, noise removal to solve this problem is an important preprocessing process in image processing. In this paper, we remove noise by using a deformable convolutional autoencoder (DeCAE) model in which deformable convolution operation is applied instead of the existing convolution operation in the convolutional autoencoder (CAE) model of deep learning. Here, the deformable convolution operation can extract features of an image in a more flexible area than the conventional convolution operation. The proposed DeCAE model has the same encoder-decoder structure as the existing CAE model, but the encoder is composed of deformable convolutional layers and the decoder is composed of conventional convolutional layers for efficient noise removal. To evaluate the performance of the DeCAE model proposed in this paper, experiments were conducted on CT images corrupted by various noises, that is, Gaussian noise, impulse noise, and Poisson noise. As a result of the performance experiment, the DeCAE model has more qualitative and quantitative measures than the traditional filters, that is, the Mean filter, Median filter, Bilateral filter and NL-means method, as well as the existing CAE models, that is, MAE (Mean Absolute Error), PSNR (Peak Signal-to-Noise Ratio) and SSIM. (Structural Similarity Index Measure) showed excellent results.

• Journal of the Korean Institute of Telematics and Electronics A
• /
• v.32A no.11
• /
• pp.52-62
• /
• 1995

We developed a two-dimensional numerical simulator which can analyze the electrical as well as optical characteristics and evaluate the detection performances of self-signal processing infrared detectors. It solves the poisson equation and the electron, hole current continuity equations including the optical generation and recombination models. To speed up convergency rate. the Newton algorithm is used. Automatic triangular grid generator make it easy to simulate the devices with the various read-out geometries. This simulator can show the variation of spatial resolution which is caused by the transit velocity and transit time dispersion in bifurcate and horn geometries respectively. Also, we calculated the responsivity, noise, and detectivity in respect of the applied electric field and background field-of-view. The results obtained from simulation correspond to those of experiments, and it is verified that horn read-out geometry has the superior spatial resolution and detection performance to bifurcate geometry.

• Journal of the Korean Institute of Telematics and Electronics
• /
• v.16 no.6
• /
• pp.22-30
• /
• 1979

A new circular ring position camera tomographic system termed "Oscillatory Dichotomic Ring" system is proposed and its performance is simulated. It is basically a circular ring system, composed of two half rings, which has the capability of scanning so that any sampling intervals can be obtained. Since finer sampling means poorer photon statistcs, simulations with varous signal dependent statistical noise effects, ray sampling and arrangement as well as related artifacts peculiar to the proposed Dichotomic Ring system are made.

• Journal of Ocean Engineering and Technology
• /
• v.22 no.4
• /
• pp.1-7
• /
• 2008

The primary aim of the paper is to solve an unstable axisymmetric initial value problem of wave propagation when given initial data that is deteriorated by noise such as measurement error. To overcome the instability of the problem, Tikhonov's regularization, known as a non-iterative numerical regularization method, is introduced to solve the problem. The L-curvecriterion is introduced to find the optimal regularization parameter for the solution. It is confirmed that fairly stable solutions are realized and that they are accurate when compared to the exact solution.

• Nuclear Medicine and Molecular Imaging
• /
• v.42 no.2
• /
• pp.118-126
• /
• 2008

Statistical image reconstruction methods have played an important role in emission computed tomography (ECT) since they accurately model the statistical noise associated with gamma-ray projection data. Although the use of statistical methods in clinical practice in early days was of a difficult problem due to high per-iteration costs and large numbers of iterations, with the development of fast algorithms and dramatically improved speed of computers, it is now inevitably becoming more practical. Some statistical methods are indeed commonly available from nuclear medicine equipment suppliers. In this paper, we first describe a mathematical background for statistical reconstruction methods, which includes assumptions underlying the Poisson statistical model, maximum likelihood and maximum a posteriori approaches, and prior models in the context of a Bayesian framework. We then review a recent progress in developing fast iterative algorithms.

• Journal of The Korean Astronomical Society
• /
• v.21 no.2
• /
• pp.155-171
• /
• 1988

We have examined how sensitively the extinction value determined by the method of star-count depends on such factors as the plate limit, the size of counting reseau, the non-linearity in the number distribution of stars with magnitude, and the angular resolution demanded by the given problem. We let the Poisson distribution portray the statistical nature of the countings, and chose the region containing the globule Barnard 361 as an example field. Uncertainties due to various combinations of the factors are presented in graphic forms: (1) Dynamic range in the extinction measurements is evaluated as a function of reseau size for varying plate limits. (2) Statistical errors involved in the star-count are analized in terms of the signal-to-noise ratio, the plate limit and the reseau size. (3) Systematic error due to the non-linearity in the number distribution are thoroughly analized. (4) Finally, a methodology is presented for correcting the systematic error in the observed radial density gradient. These graphs are meant to be used in selecting proper size of the reseau and in estimating errors inherent to the star-count analysis.

• Journal of applied mathematics & informatics
• /
• v.39 no.5_6
• /
• pp.859-882
• /
• 2021

In the proposed paper, a new algorithm based on Nonlinear Feedback Shift Register (NLFSR) and modified RC4A (Rivest Cipher 4A) cipher is introduced. NLFSR is used for image pixel scrambling while modified RC4A algorithm is used for pixel substitution. NLFSR used in this algorithm is of order 27 with maximum period 2^27-1 which was found using Field Programmable Gate Arrays (FPGA), a searching method. Modified RC4A algorithm is the modification of RC4A and is modified by introducing non-linear rotation operator in the Key Scheduling Algorithm (KSA) of RC4A cipher. Analysis of occlusion attack (up to 62.5% pixels), noise (salt and pepper, Poisson) attack and key sensitivity are performed to assess the concreteness of the proposed method. Also, some statistical and security analyses are evaluated on various images of different size to empirically assess the robustness of the proposed scheme.