• The Journal of the Acoustical Society of Korea
• /
• v.40 no.2
• /
• pp.130-138
• /
• 2021

This paper presents an effective structure by applying various normalization to Convolutional Neural Networks (CNN) for seismic event classification. Normalization techniques can not only improve the learning speed of neural networks, but also show robustness to noise. In this paper, we analyze the effect of input data normalization and hidden layer normalization on the deep learning model for seismic event classification. In addition an effective model is derived through various experiments according to the structure of the applied hidden layer. As a result of various experiments, the model that applied input data normalization and weight normalization to the first hidden layer showed the most stable performance improvement.

• Journal of Information Processing Systems
• /
• v.16 no.6
• /
• pp.1391-1406
• /
• 2020

A zero-watermarking algorithm in transform domain based on RGB channel and voting strategy is proposed. The registration and identification of ownership have achieved copyright protection for color images. In the ownership registration, discrete wavelet transform (DWT), discrete cosine transform (DCT), and singular value decomposition (SVD) are used comprehensively because they have the characteristics of multi-resolution, energy concentration and stability, which is conducive to improving the robustness of the proposed algorithm. In order to take full advantage of the characteristics of the image, we use three channels of R, G, and B of a color image to construct three master shares, instead of using data from only one channel. Then, in order to improve security, the master share is superimposed with the copyright watermark encrypted by the owner's key to generate an ownership share. When the ownership is authenticated, copyright watermarks are extracted from the three channels of the disputed image. Then using voting decisions, the final copyright information is determined by comparing the extracted three watermarks bit by bit. Experimental results show that the proposed zero watermarking scheme is robust to conventional attacks such as JPEG compression, noise addition, filtering and tampering, and has higher stability in various common color images.

• Journal of Broadcast Engineering
• /
• v.26 no.7
• /
• pp.844-854
• /
• 2021

Semi-supervised learning (SSL) and few-shot learning (FSL) have shown impressive performance even then the volume of labeled data is very limited. However, SSL and FSL can encounter a significant performance degradation if the diversity gap between the labeled and unlabeled data is high. To reduce this diversity gap, we propose a novel scheme that relies on an autoencoder for generating pseudo examples. Specifically, the autoencoder is trained on a specific class using the available labeled data and the decoder of the trained autoencoder is then used to generate N samples of that specific class based on N random noise, sampled from a standard normal distribution. The above process is repeated for all the classes. Consequently, the generated data reduces the diversity gap and enhances the model performance. Extensive experiments on MNIST and FashionMNIST datasets for SSL and FSL verify the effectiveness of the proposed approach in terms of classification accuracy and robustness against adversarial attacks.

• Journal of Positioning, Navigation, and Timing
• /
• v.11 no.3
• /
• pp.147-156
• /
• 2022

In this paper, we propose Time-Division-Multiplexing Tertiary Offset Carrier (TDMTOC), a novel GNSS modulation based on Tertiary Offset Carrier (TOC) modulation. The TDMTOC modulation multiplexes two three-level signals (i.e., -1, 0, and 1) while crossing over time, and is a type of TOC modulation designed specifically for signal multiplexing. The proposed modulation generates TDMTOC subcarriers of two different phases by simply combining two Binary Offset Carrier (BOC) subcarriers by addition or subtraction. TDMTOC has better correlation and spectral properties than conventional BPSK, BOC, and MBOC modulation techniques, and has good power and spectral efficiency since it can multiplex signals without power loss similar to time division multiplexing. To prove this, we introduce the multiplexing process of TDMTOC, and compare TDMTOC with Binary Phase Shift Keying (BPSK), BOC, Composite BOC (CBOC), and Time Multiplexed BOC (TMBOC) that are currently serviced in GNSS by simulations of various aspects. Through the simulation results, we prove that TDMTOC has better correlation property than modulations currently used in GNSS, less intersystem interference due to its wide spectrum property, and robustness in multipath and noise channel environments.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.18 no.1
• /
• pp.77-84
• /
• 2023

In autonomous robots, obstacle avoidance is a key feature. Potential Field is the most widely used method in this field. Such method requires real-time calculation of the nearest point of the obstacle from the robot, which involves difficulty of reliably segmenting the obstacle region from the distance sensor data profile. In this paper, Active Min-Depth Filter is introduced to obtain the nearest point of each obstacle using real-time calculation but without segmentation. Through simulations on various sensor noise environments, the robustness of the Active Min-Depth Filter could be confirmed, and successful results were obtained by applying real-world moving robots.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2021.11a
• /
• pp.700-703
• /
• 2021

Infrared cameras are widely used in recent research for automatic monitoring the abnormal behaviors of the pig. However, when deployed in real pig farms, infrared cameras always get polluted due to the harsh environment of pig farms which negatively affects the performance of pig monitoring. In this paper, we propose a real-time noise-robust infrared camera-based pig automatic monitoring system to improve the robustness of pigs' automatic monitoring in real pig farms. The proposed system first uses a preprocessor with a U-Net architecture that was trained as a GAN generator to transform the noisy images into clean images, then uses a YOLOv5-based detector to detect pigs. The experimental results show that with adding the preprocessing step, the average pig detection precision improved greatly from 0.639 to 0.759.

• Journal of the Korean Society of Visualization
• /
• v.21 no.2
• /
• pp.14-23
• /
• 2023

The pressure difference across stenotic blood vessels is a commonly used clinical metric for diagnosing many cardiovascular diseases. At present, most clinical pressure measurements rely solely on invasive catheterization. In this study, we propose a novel method for non-invasive pressure estimation using the incompressible Navier-Stokes equations and a 3D multi-path integration approach. We verify spatio-temporal convergence on an in-silico dataset of a cylindrical straight pipe phantom with steady and pulsatile flow fields. We then evaluate the proposed method on an in vitro dataset of reconstructed control, pre-operative, and post-operative carotid artery cases acquired from 4D flow MRI. The performance of our method is compared to existing approaches based on the pressure Poisson equation and work-energy relative pressure. The results demonstrate the proposed method's high accuracy, robustness to spatio-temporal subsampling, and reduced sensitivity to noise, highlighting its great potential for non-invasive pressure estimation.

• International Journal of Computer Science & Network Security
• /
• v.24 no.3
• /
• pp.93-100
• /
• 2024

In this paper, we propose a novel robust blind crypto-watermarking method for medical images security based on hiding of DICOM patient information (patient name, age...) in the medical imaging. The DICOM patient information is encrypted using the AES standard algorithm before its insertion in the medical image. The cover image is divided in blocks of 8x8, in each we insert 1-bit of the encrypted watermark in the hybrid transform domain by applying respectively the 2D-LWT (Lifting wavelet transforms), the 2D-DCT (discrete cosine transforms), and the SVD (singular value decomposition). The scheme is tested by applying various attacks such as noise, filtering and compression. Experimental results show that no visible difference between the watermarked images and the original images and the test against attack shows the good robustness of the proposed algorithm.

• The Journal of the Acoustical Society of Korea
• /
• v.29 no.2
• /
• pp.97-101
• /
• 2010

FIR type system identification with noisy input and output data can be solved by a total least squares (TLS) estimation. However, the performance of the TLS estimation is very sensitive to the ratio between the variances of the input and output noises. In this paper, we propose an iterative convex combination algorithm between TLS and least squares (LS). This combined algorithm shows robustness against the noise variance ratio. Consequently, the practical workability of the TLS method with noisy data has been significantly broadened.

• Asia pacific journal of information systems
• /
• v.29 no.4
• /
• pp.789-816
• /
• 2019

Text classification is a challenging task, especially when dealing with a huge amount of text data. The performance of a classification model can be varied depending on what type of words contained in the document corpus and what type of features generated for classification. Aside from proposing a new modified version of the existing algorithm or creating a new algorithm, we attempt to modify the use of data. The classifier performance is usually affected by the quality of learning data as the classifier is built based on these training data. We assume that the data from different domains might have different characteristics of noise, which can be utilized in the process of learning the classifier. Therefore, we attempt to enhance the robustness of the classifier by injecting the heterogeneous data artificially into the learning process in order to improve the classification accuracy. Semi-supervised approach was applied for utilizing the heterogeneous data in the process of learning the document classifier. However, the performance of document classifier might be degraded by the unlabeled data. Therefore, we further proposed an algorithm to extract only the documents that contribute to the accuracy improvement of the classifier.