• Journal of the Korean Society for Precision Engineering
• /
• v.19 no.1
• /
• pp.158-164
• /
• 2002

This paper proposes the optimization of ultrasonic signal recognition in ferrite carbon steel weld zone using neural networks. For these purposes, the ultrasonic signals for defects as porosity, incomplete penetration and slag inclusion in the weld zone are acquired in the type of time series data. And then their applications evaluated feature extraction based on the time-frequency-attractor domain(peak to peak, rise time, rise slope, fall time, fall slope, pulse duration, power spectrum, and bandwidth) and attractor characteristics (fractal dimension and attractor quadrant) etc. The proposed neural networks system in this study can enhances performance of ultrasonic signal recognition.

• Journal of Korea Society of Digital Industry and Information Management
• /
• v.4 no.2
• /
• pp.57-66
• /
• 2008

There have been proposed various copyright protection protocols in network-based digital multimedia distribution framework. However, most of conventional copyright protection protocols are focused on the stability of copyright information embedding/extracting and the access control to data suitable for user's authority but overlooked the privacy of copyright owner and user in authentication process of copyright and access information. In this paper, we propose a solution that builds a privacy-preserving proof of copyright ownership of digital contents in conjunction with keyword search scheme. The appeal of our proposal is three-fold: (1) content providers maintain stable copyright ownership in the distribution of digital contents; (2) the proof process of digital contents ownership is very secure in the view of preserving privacy; (3) the proposed protocol is the copyright protection protocol added by indexing process but is balanced privacy and efficiency concerns for its practical use.

• Proceedings of the IEEK Conference
• /
• 2002.06d
• /
• pp.319-322
• /
• 2002

In this paper, we present a digital authentication technique using content-based watermarking in digital images. To digest the image contents, Hopfield network is employed on the block-based edge image. The Hopfield function extracts the same tit fur similarly looking blocks so that the values are unlikely to change to the innocuous manipulations while being changed far malicious manipulations. By inputting the extracted bit sequence with secret key to the cryptographic hash function, we generate a watermark for each block by seeding a pseudo random number generator with a hash output Therefore, the proposed authentication technique can distinguish between malicious attacks and innocuous attacks. Watermark embedding is based on the block-based spread spectrum method in DCT domain and the strength of watermark is adjusted according to the local statistics of DCT coefficients in a zig-zag scan line in AC subband. The numerical experiments show that the proposed technique is very efficient in the performance of robust authentication.

• Journal of Communications and Networks
• /
• v.14 no.2
• /
• pp.213-221
• /
• 2012

The architecture of an interconnection network is usually represented by a graph, and a graph G is bipancyclic if it contains a cycle for every even length from 4 to ${\mid}V(G){\mid}$. In this article, we analyze the conditional edge-fault-tolerant properties of an enhanced hypercube, which is an attractive variant of a hypercube that can be obtained by adding some complementary edges. For any n-dimensional enhanced hypercube with at most (2n-3) faulty edges in which each vertex is incident with at least two fault-free edges, we showed that there exists a fault-free cycle for every even length from 4 to $2^n$ when n($n{\geq}3$) and k have the same parity. We also show that a fault-free cycle for every odd length exists from n-k+2 to $2^n-1$ when n($n{\geq}2$) and k have the different parity.

• Annual Conference on Human and Language Technology
• /
• 2015.10a
• /
• pp.157-162
• /
• 2015

자동통역(Speech-to-speech translation)의 최우선 단계인 음성인식과정에서 발생한 오류문장은 대부분 비문법적 구조를 갖거나 의미를 이해할 수 없는 문장들이다. 이러한 문장으로 자동번역을 할 경우 심각한 통역오류가 발생하게 되어 이에 대한 개선이 반드시 필요한 상황이다. 이에 본 논문에서는 음성인식 오류문장이 정상적인 인식문장에 비해 비문법적이거나 무의미하다는 특징을 이용하여 DNN(Deep Neural Network) 기반 음성인식오류 판별기를 구현하였으며 84.20%의 오류문장 분류성능결과를 얻었다.

• Annual Conference of KIPS
• /
• 2018.10a
• /
• pp.763-766
• /
• 2018

표정은 사람들 사이에서 감정을 표현하는 강력한 비언어적 수단이다. 표정 인식은 기계학습에서 아주 중요한 분야 중에 하나이다. 표정 인식에 사용되는 기계학습 모델들은 사람 수준의 성능을 보여준다. 하지만 좋은 성능에도 불구하고, 기계학습 모델들은 표정 인식 결과에 대한 근거나 설명을 제공해주지 못한다. 이 연구는 표정 인식의 근거로서 Facial Action Coding Unit(AUs)을 사용하기 위해서 CK+ Dataset을 사용하여 표정 인식을 학습한 Convolutional Neural Network(CNN) 모델이 추출한 특징들을 t-distributed stochastic neighbor embedding(t-SNE)을 사용하여 시각화한 뒤, 인식된 표정과 AUs 사이의 분포의 연관성을 확인하는 연구이다.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.9 no.10
• /
• pp.4143-4159
• /
• 2015

We-the-media (aka wedia) is a concept where the users of social networking sites, such as Facebook, turn into the broadcasters. This study used the popular application Bitstrips as the experiment tool. Facebook was used as the Wedia platform for publishing designed comics, then used the three elements of Goffman's dramaturgy model-role, scene and dialog-to analyze 265 comics created by 3 researchers and observe the audience's responses within 9 months. The results showed that people want to see a good story with positive dialogue, and prefer scene is school more than work. As all these elements are controllable, Wedia communication has the potential for more applications. We also found that including the elements of news, gambling and gift-giving tended to trigger greater response. Furthermore, We suggesting that such embedding of product information in web episodes (webisodes) with caricature could be a successful marketing strategy.

• Journal of Information Processing Systems
• /
• v.15 no.1
• /
• pp.169-180
• /
• 2019

Transmitting visual information over a broadcasting network is not only prone to a copyright violation but also is a forgery. Authenticating such information and protecting its authorship rights call for more advanced data encoding. To this end, electronic watermarking is often adopted to embed inscriptive signature in imaging data. Most existing watermarking methods while focusing on robustness against degradation remain lacking of measurement against security loophole in which the encrypting scheme once discovered may be recreated by an unauthorized party. This could reveal the underlying signature which may potentially be replaced or forged. This paper therefore proposes a novel digital watermarking scheme in temporal-frequency domain. Unlike other typical wavelet based watermarking, the proposed scheme employed the Lorenz chaotic map to specify embedding positions. Effectively making this is not only a formidable method to decrypt but also a stronger will against deterministic attacks. Simulation report herein highlights its strength to withstand spatial and frequent adulterations, e.g., lossy compression, filtering, zooming and noise.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.14 no.11
• /
• pp.4557-4572
• /
• 2020

The Alaska competition provides an opportunity to study the practical problems of real-world steganalysis. Participants are required to solve steganalysis involving various embedding schemes, inconsistency JPEG Quality Factor and various processing pipelines. In this paper, we propose a method to ensemble multiple deep learning steganalyzers. We select SRNet and RESDET as our base models. Then we design a three-layers model ensemble network to fuse these base models and output the final prediction. By separating the three colors channels for base model training and feature replacement strategy instead of simply merging features, the performance of the model ensemble is greatly improved. The proposed method won second place in the Alaska 1 competition in the end.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.16 no.1
• /
• pp.305-318
• /
• 2022

Abusive behaviors have become a common issue in many online social media platforms. Profanity is common form of abusive behavior in online. Social media platforms operate the filtering system using popular profanity words lists, but this method has drawbacks that it can be bypassed using an altered form and it can detect normal sentences as profanity. Especially in Korean language, the syllable is composed of graphemes and words are composed of multiple syllables, it can be decomposed into graphemes without impairing the transmission of meaning, and the form of a profane word can be seen as a different meaning in a sentence. This work focuses on the problem of filtering system mis-detecting normal phrases with profane phrases. For that, we proposed the deep learning-based framework including grapheme and syllable separation-based word embedding and appropriate CNN structure. The proposed model was evaluated on the chatting contents from the one of the famous online games in South Korea and generated 90.4% accuracy.