• 대한임베디드공학회논문지
• /
• 제3권1호
• /
• pp.8-18
• /
• 2008

In this paper, we propose an efficient algorithm using wavelet transform for still image database retrieval. Especially, it uses only the lowest frequency sub-band in multi-level wavelet transform so that a retrieval system uses a smaller quantity of memory and takes a faster processing time. We extract different textured features, statistical information such as mean, variance and histogram, from low frequency sub-band. Then we measure the distances between the query image and the images in a database in terms of these features. To obtain good retrieval performance, we use the first feature (mean and variance of wavelet coefficients) to filter out most of the unlikely images. The rest of the images are considered to be candidate images. Then we apply the second feature (histogram of wavelet coefficient) to rank all the candidate images. To evaluate the algorithm, we create various distorted image databases using MIT VisTex texture images and PICS natural images. Through simulations, we demonstrate that our method can achieve performance satisfactorily in terms of the retrieval accuracy as well as the both memory requirement and computational complexity. Therefore it is expected to provide good retrieval solution for JPEG-2000 using wavelet transform.

• 한국멀티미디어학회논문지
• /
• 제20권5호
• /
• pp.769-781
• /
• 2017

Deep learning networks like Convolutional Neural Networks (CNNs) show successful performances in many computer vision applications such as image classification, object detection, and so on. For implementation of deep learning networks in embedded system with limited processing power and memory, deep learning network may need to be simplified. However, simplified deep learning network cannot learn every possible scene. One realistic strategy for embedded deep learning network is to construct a simplified deep learning network model optimized for the scene images of the installation place. Then, automatic training will be necessitated for commercialization. In this paper, as an intermediate step toward automatic training under fisheye camera environments, we study more precise human localization in fisheye images, and propose an accurate human localization method, Automatic Ground-Truth Labelling Method (AGTLM). AGTLM first localizes candidate human object bounding boxes by utilizing GoogLeNet-LSTM approach, and after reassurance process by GoogLeNet-based CNN network, finally refines them more correctly and precisely(tightly) by applying saliency object detection technique. The performance improvement of the proposed human localization method, AGTLM with respect to accuracy and tightness is shown through several experiments.

• 한국지진공학회:학술대회논문집
• /
• 한국지진공학회 2003년도 춘계 학술발표회논문집
• /
• pp.329-336
• /
• 2003

The box system that is composed only of reinforced concrete walls and slabs are adopted on many high-rise apartment buildings recently constructed in Korea. And the framed structure with shear wall core that can effectively resist horizontal forces is frequently adopted for the structural system for high-rise building structures. In these structures, a shear wall may have one or more openings for functional reasons. It is necessary to use subdivided finite elements for accurate analysis of the shear wall with openings. But it would take significant amount of computational time and memory if the entire building structure is subdivided into a finer mesh. An efficient analysis method that can be used regardless of the number, size and location of openings is proposed in this study. The analysis method uses super element, substructure, matrix condensation technique and fictitious beam technique. Three-dimensional analyses of the box system and the framed structure with shear wall core having various types of openings were peformed to verify the efficiency of the proposed method. It was confirmed that the proposed method have outstanding accuracy with drastically reduced time and computer memory from the analyses of example structures.

• 한국지진공학회:학술대회논문집
• /
• 한국지진공학회 2003년도 춘계 학술발표회논문집
• /
• pp.50-58
• /
• 2003

We have developed a locally variable time-step scheme matching with discontinuous grids in the flute-difference method for the efficient simulation of seismic wave propagation. The first-order velocity-stress formulations are used to obtain the spatial derivatives using finite-difference operators on a staggered grid. A three-times coarser grid in the high-velocity region compared with the grid in the low-velocity region is used to avoid spatial oversampling. Temporal steps corresponding to the spatial sampling ratio between both regions are determined based on proper stability criteria. The wavefield in the margin of the region with smaller time-step are linearly interpolated in time using the values calculated in the region with larger one. The accuracy of the proposed scheme is tested through comparisons with analytic solutions and conventional finite-difference scheme with constant grid spacing and time step. The use of the locally variable time-step scheme with discontinuous grids results in remarkable saving of the computation time and memory requirement with dependency of the efficiency on the simulation model. This implies that ground motion for a realistic velocity structures including near-surface sediments can be modeled to high frequency (several Hz) without requiring severe computer memory

• Journal of Information Processing Systems
• /
• 제17권4호
• /
• pp.707-720
• /
• 2021

To improve the accuracy of the recommendation system, multi-criteria recommendation systems have been widely researched. However, it is highly complicated to extract the preferred features of users and items from the data. To this end, subjective indicators, which indicate a user's priorities for personalized recommendations, should be derived. In this study, we propose a method for generating recommendation candidates by predicting multi-criteria ratings from reviews and using them to derive user priorities. Using a deep learning model based on convolutional neural network (CNN) and bidirectional long short-term memory (BiLSTM), multi-criteria prediction ratings were derived from reviews. These ratings were then aggregated to form a linear regression model to predict the overall rating. This model not only predicts the overall rating but also uses the training weights from the layers of the model as the user's priority. Based on this, a new score matrix for recommendation is derived by calculating the similarity between the user and the item according to the criteria, and an item suitable for the user is proposed. The experiment was conducted by collecting the actual "TripAdvisor" dataset. For performance evaluation, the proposed method was compared with a general recommendation system based on singular value decomposition. The results of the experiments demonstrate the high performance of the proposed method.

• 한국기계가공학회지
• /
• 제20권12호
• /
• pp.85-91
• /
• 2021

In our daily life, artificial intelligence performs simple and complicated tasks like us, including operating mobile phones and working at homes and workplaces. Artificial intelligence is used in industrial technology for diagnosing various types of equipment using the machine learning technology. This study presents a fault mode effect analysis (FMEA) of start motors using machine learning and big data. Through multiple data collection, we observed that the primary failure of the start motor was caused by the melting of the magnetic switch inside the start motor causing it to fail. Long-short-term memory (LSTM) was used to diagnose the condition of the magnetic locations, and synthetic data were generated using the synthetic minority oversampling technique (SMOTE). This technique has the advantage of increasing the data accuracy. LSTM can also predict a start motor failure.

• 한국산업융합학회 논문집
• /
• 제24권6_1호
• /
• pp.693-698
• /
• 2021

As technology advances, there is increasing need for research in different fields where this technology is applied. On of the most researched topic in computer vision is Human activity recognition (HAR), which has widely been implemented in various fields which include healthcare, video surveillance and education. We therefore present in this paper a human activity recognition system based on scale and rotation while employing the Kinect depth sensors to obtain the human skeleton joints. In contrast to previous approaches that use joint angles, in this paper we propose that each limb has an angle with the X, Y, Z axes which we employ as feature vectors. The use of the joint angles makes our system scale invariant. We further calculate the body relative direction in the egocentric coordinates in order to provide the rotation invariance. For the system parameters, we employ 8 limbs with their corresponding angles each having the X, Y, Z axes from the coordinate system as feature vectors. The extracted features are finally trained and tested with the Long short term memory (LSTM) Network which gives us an average accuracy of 98.3%.

• Journal of Information Processing Systems
• /
• 제18권5호
• /
• pp.614-627
• /
• 2022

Aspect-based sentiment analysis is to discover the sentiment polarity towards an aspect from user-generated natural language. So far, most of the methods only use the implicit position information of the aspect in the context, instead of directly utilizing the position relationship between the aspect and the sentiment terms. In fact, neighboring words of the aspect terms should be given more attention than other words in the context. This paper studies the influence of different position embedding methods on the sentimental polarities of given aspects, and proposes a position embedding interactive attention network based on a long short-term memory network. Firstly, it uses the position information of the context simultaneously in the input layer and the attention layer. Secondly, it mines the importance of different context words for the aspect with the interactive attention mechanism. Finally, it generates a valid representation of the aspect and the context for sentiment classification. The model which has been posed was evaluated on the datasets of the Semantic Evaluation 2014. Compared with other baseline models, the accuracy of our model increases by about 2% on the restaurant dataset and 1% on the laptop dataset.

• 디지털산업정보학회논문지
• /
• 제18권4호
• /
• pp.67-75
• /
• 2022

Recently, e-learning has been attracting significant attention due to COVID-19. However, while e-learning has many advantages, it has disadvantages as well. One of the main disadvantages of e-learning is that it is difficult for teachers to continuously and systematically monitor learners. Although services such as personalized e-learning are provided to compensate for the shortcoming, systematic monitoring of learners' concentration is insufficient. This study suggests a method to evaluate the learner's concentration by applying machine learning techniques. In this study, emotion and gaze data were extracted from 184 videos of 92 participants. First, the learners' concentration was labeled by experts. Then, statistical-based status indicators were preprocessed from the data. Random Forests (RF), Support Vector Machines (SVMs), Multilayer Perceptron (MLP), and an ensemble model have been used in the experiment. Long Short-Term Memory (LSTM) has also been used for comparison. As a result, it was possible to predict e-learners' concentration with an accuracy of 90.54%. This study is expected to improve learners' immersion by providing a customized educational curriculum according to the learner's concentration level.

• International Journal of Computer Science & Network Security
• /
• 제21권11호
• /
• pp.73-80
• /
• 2021

Nowadays microblogs have become the most popular platforms to obtain and spread information. Twitter is one of the most used platforms to share everyday life event. However, rumors and misinformation on Arabic social media platforms has become pervasive which can create inestimable harm to society. Therefore, it is imperative to tackle and study this issue to distinguish the verified information from the unverified ones. There is an increasing interest in rumor detection on microblogs recently, however, it is mostly applied on English language while the work on Arabic language is still ongoing research topic and need more efforts. In this paper, we propose a combined Convolutional Neural Network (CNN) and Long Short-Term Memory (LSTM) to detect rumors on Twitter dataset. Various experiments were conducted to choose the best hyper-parameters tuning to achieve the best results. Moreover, different neural network models are used to evaluate performance and compare results. Experiments show that the CNN-LSTM model achieved the best accuracy 0.95 and an F1-score of 0.94 which outperform the state-of-the-art methods.