• Journal of Korea Multimedia Society
• /
• v.14 no.5
• /
• pp.595-606
• /
• 2011

This paper proposes a robust watermarking scheme for GIS digital map by using the geometric properties of polyline and polygon, which are the fundamental components in vector data structure. In the proposed scheme, we calculate the lengths and the perimeters of all polylines and polygons in a map and cluster them to a number of groups. Then we embed the binary watermark by changing the mean of lengths and perimeters in an embedding group. For improving the safety and robustness, we permute the binary watermark through PRNS(pseudo-random number sequence) processing and embed it repeatedly in a model. Experimental results verified that our scheme has a good invisibility, safety and robustness to various geometric attacks and also our scheme needs not the original map in the extracting process of watermark.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.16 no.10
• /
• pp.3211-3229
• /
• 2022

To solve the problems existing in the process of Weibo disaster rumor recognition, such as lack of corpus, poor text standardization, difficult to learn semantic information, and simple semantic features of disaster rumor text, this paper takes Sina Weibo as the data source, constructs a dataset for Weibo disaster rumor recognition, and proposes a deep learning model BERT_AT_Stacked LSTM for Weibo disaster rumor recognition. First, add adversarial disturbance to the embedding vector of each word to generate adversarial samples to enhance the features of rumor text, and carry out adversarial training to solve the problem that the text features of disaster rumors are relatively single. Second, the BERT part obtains the word-level semantic information of each Weibo text and generates a hidden vector containing sentence-level feature information. Finally, the hidden complex semantic information of poorly-regulated Weibo texts is learned using a Stacked Long Short-Term Memory (Stacked LSTM) structure. The experimental results show that, compared with other comparative models, the model in this paper has more advantages in recognizing disaster rumors on Weibo, with an F1_Socre of 97.48%, and has been tested on an open general domain dataset, with an F1_Score of 94.59%, indicating that the model has better generalization.

• Journal of the Korean Mathematical Society
• /
• v.42 no.1
• /
• pp.129-134
• /
• 2005

Lpt X be a reduced Stein space and L a holomorphic line bundle on X. L is spanned by its global sections and the associated holomorphic map $h_L\;:\;X{\to}P(H^0(X, L)^{\ast})$ is an embedding. Choose any locally convex vector topology ${\tau}\;on\;H^0(X, L)^{\ast}$ stronger than the weak-topology. Here we prove that $h_L(X)$ is sequentially closed in $P(H^0(X, L)^{\ast})$ and arithmetically Cohen -Macaulay. i.e. for all integers $k{\ge}1$ the restriction map ${\rho}_k\;:\;H^0(P(H^0(X, L)^{\ast}),\;O_{P(H^0(X, L)^{\ast})}(k)){\to}H^0(h_L(X),O_{hL_(X)}(k)){\cong}H^0(X, L^{\otimes{k}})$ is surjective.

• Transactions on Electrical and Electronic Materials
• /
• v.16 no.5
• /
• pp.250-253
• /
• 2015

A novel implantable CPW fed Apollian shaped antenna embedded into human tissue is proposed for ISM band biomedical applications. The proposed antenna is made compatible for implantation by embedding it in an alumina ceramic substrate(ε_r=9.8 and thickness= 0.65 mm). The proposed antenna covers the ISM band of 2.45 GHz. The radiation parameters such as return loss, xy-plane, xz-plane, and yz-plane etc., are measured and analyzed using the agilent vector network analyzer. The proposed antenna has substantial advantages, including low profile, miniaturization ability, lower return loss, better impedance matching, and high gain over conventional implanted antennas.

• Journal of Computing Science and Engineering
• /
• v.11 no.1
• /
• pp.1-8
• /
• 2017

Naive Bayes nearest neighbor (NBNN) is a simple image classifier based on identifying nearest neighbors. NBNN uses original image descriptors (e.g., SIFTs) without vector quantization for preserving the discriminative power of descriptors and has a powerful generalization characteristic. However, it has a distinct disadvantage. Its memory requirement can be prohibitively high while processing a large amount of data. To deal with this problem, we apply a spherical hashing binary code embedding technique, to compactly encode data without significantly losing classification accuracy. We also propose using an inverted index to identify nearest neighbors among binarized image descriptors. To demonstrate the benefits of our method, we apply our method to two existing NBNN techniques with an image dataset. By using 64 bit length, we are able to reduce memory 16 times with higher runtime performance and no significant loss of classification accuracy. This result is achieved by our compact encoding scheme for image descriptors without losing much information from original image descriptors.

• Korean Journal of Computational Design and Engineering
• /
• v.19 no.1
• /
• pp.68-79
• /
• 2014

Human model simulation has been widely used in various industrial areas such as ergonomic design, product evaluation and characteristic analysis of work-related musculoskeletal disorders. However, the process of building digital human models and capturing their behaviors requires many costly and time-consuming fabrication iterations. To overcome the limitations of this expensive and time-consuming process, many studies have recently presented a markerless motion capture approach that reconstructs the time-varying skeletal motions from optical devices. However, the drawback of the markerless motion capture approach is that the phenomenon of occlusion of motion data occurs in real-time human simulation. In this study, we propose a systematic method of discriminating missing or inaccurate motion data due to motion occlusion and interpolating a sequence of motion frames captured by a markerless depth camera.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2020.05a
• /
• pp.451-454
• /
• 2020

딥러닝 기술을 이용한 정치적 성향의 편향성 분류를 위하여 신문 뉴스 기사를 수집하고, 머신러닝을 위한 학습 데이터를 구축하였다. 학습 데이터의 구축은 보수 성향과 진보 성향을 대표하는 6개 언론사의 뉴스에서 정치적 성향을 이진 분류 데이터로 구축하였다. 뉴스 기사의 수집 방법으로 최근 이슈들 중에서 정치적 성향과 밀접하게 관련이 있는 키워드 15개를 선정하고 이에 관한 뉴스 기사들을 수집하였다. 그 결과로 11,584개의 학습 및 실험용 데이터를 구축하였으며, 정치적 편향성 분류를 위한 머신러닝 모델을 설계하였다. 머신러닝 기법으로 학습 및 실험을 위해 형태소 단위의 임베딩을 이용하여 문장 및 문서 임베딩으로 확장하였으며, SVM(Support Vector Machine)을 이용하여 정치적 편향성 분류 실험을 수행한 결과로 75%의 정확도를 달성하였다.

• Journal of Korea Multimedia Society
• /
• v.10 no.11
• /
• pp.1427-1438
• /
• 2007

Video watermarking technologies are classified into types of four kinds. The first type is to embed the watermark into a raw video signal and to code the watermarked video signal. Most of video watermarking technologies fall into the category of this type. The second type is to apply watermarking to the coding process, such as block DCT and quantization. The third is to directly embed the watermark into the compressed bitstream itself. Generally, it is referred as labelling rather than watermarking. Finally, the fourth is to embed the water mark into MPEG motion vector. This type has the difficulty in real-time process because of the high complexity and has the blocking effects because of DCT-based on coder. In this paper, we proposed the digital video watermarking that embed the watermark in SPIHT video code for I-frame using motion vector analysis. This method can remove the blocking effect occurred at the DCT-based on coder and obtain video data that has progressive transmission property. The proposed method is to select the region for the watermark embedding in I frame using motion vector estimated from the previous P or B frame. And then, it is to perform DWT and embed the watermark based on HVS into the wavelet coefficients in the same subband of DWT as the motion vector direction. Finally, the watermarked video bitstream is obtained by the SPIHT coder. The experimental results verified that the proposed method has the invisibility from the objective and subjective image quality and the robustness against the various SPIHT compression and MPEG re-code.

• Journal of the Institute of Electronics Engineers of Korea SD
• /
• v.42 no.4 s.334
• /
• pp.45-52
• /
• 2005

This paper presents the characterization of through hole vias on printed circuit board (PCB) through the time domain and frequency domain measurements. The time domain measurement was performed on a single via using the TDR, and the model parameters were extracted by the fitting simulation using HSPICE. The frequency domain measurement was also performed by using 2 port VNA, and the model parameters were extracted by fitting simulation with ADS. Using the ABCD matrices, the do-embedding equations were derived probing in the same plane in the VNA measurement. Based on the single via characterization, the differential via characterization was also performed by using TDR measurements. The time domain measurements were performed by using the odd mode and even mode sources in TDR module, and the Parameter values were extracted by fitting with HSPICE. Comparing measurements with simulations, the maximum calculated differences were $14\%$ for single vias and $17\%$ for differential vias.

• Smart Media Journal
• /
• v.10 no.1
• /
• pp.39-46
• /
• 2021

This study is about a method of extracting a summary from a news article in consideration of the importance of each sentence constituting the article. We propose a method of calculating sentence importance by extracting the probabilities of topic sentence, similarity with article title and other sentences, and sentence position as characteristics that affect sentence importance. At this time, a hypothesis is established that the Topic Sentence will have a characteristic distinct from the general sentence, and a deep learning-based classification model is trained to obtain a topic sentence probability value for the input sentence. Also, using the pre-learned ELMo language model, the similarity between sentences is calculated based on the sentence vector value reflecting the context information and extracted as sentence characteristics. The topic sentence classification performance of the LSTM and BERT models was 93% accurate, 96.22% recall, and 89.5% precision, resulting in high analysis results. As a result of calculating the importance of each sentence by combining the extracted sentence characteristics, it was confirmed that the performance of extracting the topic sentence was improved by about 10% compared to the existing TextRank algorithm.