• Journal of Intelligence and Information Systems
• /
• v.25 no.1
• /
• pp.163-177
• /
• 2019

As smartphones are getting widely used, human activity recognition (HAR) tasks for recognizing personal activities of smartphone users with multimodal data have been actively studied recently. The research area is expanding from the recognition of the simple body movement of an individual user to the recognition of low-level behavior and high-level behavior. However, HAR tasks for recognizing interaction behavior with other people, such as whether the user is accompanying or communicating with someone else, have gotten less attention so far. And previous research for recognizing interaction behavior has usually depended on audio, Bluetooth, and Wi-Fi sensors, which are vulnerable to privacy issues and require much time to collect enough data. Whereas physical sensors including accelerometer, magnetic field and gyroscope sensors are less vulnerable to privacy issues and can collect a large amount of data within a short time. In this paper, a method for detecting accompanying status based on deep learning model by only using multimodal physical sensor data, such as an accelerometer, magnetic field and gyroscope, was proposed. The accompanying status was defined as a redefinition of a part of the user interaction behavior, including whether the user is accompanying with an acquaintance at a close distance and the user is actively communicating with the acquaintance. A framework based on convolutional neural networks (CNN) and long short-term memory (LSTM) recurrent networks for classifying accompanying and conversation was proposed. First, a data preprocessing method which consists of time synchronization of multimodal data from different physical sensors, data normalization and sequence data generation was introduced. We applied the nearest interpolation to synchronize the time of collected data from different sensors. Normalization was performed for each x, y, z axis value of the sensor data, and the sequence data was generated according to the sliding window method. Then, the sequence data became the input for CNN, where feature maps representing local dependencies of the original sequence are extracted. The CNN consisted of 3 convolutional layers and did not have a pooling layer to maintain the temporal information of the sequence data. Next, LSTM recurrent networks received the feature maps, learned long-term dependencies from them and extracted features. The LSTM recurrent networks consisted of two layers, each with 128 cells. Finally, the extracted features were used for classification by softmax classifier. The loss function of the model was cross entropy function and the weights of the model were randomly initialized on a normal distribution with an average of 0 and a standard deviation of 0.1. The model was trained using adaptive moment estimation (ADAM) optimization algorithm and the mini batch size was set to 128. We applied dropout to input values of the LSTM recurrent networks to prevent overfitting. The initial learning rate was set to 0.001, and it decreased exponentially by 0.99 at the end of each epoch training. An Android smartphone application was developed and released to collect data. We collected smartphone data for a total of 18 subjects. Using the data, the model classified accompanying and conversation by 98.74% and 98.83% accuracy each. Both the F1 score and accuracy of the model were higher than the F1 score and accuracy of the majority vote classifier, support vector machine, and deep recurrent neural network. In the future research, we will focus on more rigorous multimodal sensor data synchronization methods that minimize the time stamp differences. In addition, we will further study transfer learning method that enables transfer of trained models tailored to the training data to the evaluation data that follows a different distribution. It is expected that a model capable of exhibiting robust recognition performance against changes in data that is not considered in the model learning stage will be obtained.

• Proceedings of the Korean Institute of Electrical and Electronic Material Engineers Conference
• /
• 2008.11a
• /
• pp.122-122
• /
• 2008

Recently, Metal/Alumina/Silicon-Nitride/Silicon-Oxide/Silicon (MANOS) structures are one of the most attractive candidates to realize vertical scaling of high-density NAND flash memory [1]. However, as ANO layers are miniaturized, negative and positive bias temperature instability (NBTI/PBTI), such as the flat band voltage shift, ${\Delta}V_{FB}$, the interfacial trap density increase, ${\Delta}D_{it}$, the gate leakage current, ${\Delta}I_G$. and the retention characteristics, in MONOS capacitors, becomes an important issue in terms of reliability. It is well known that tunnel oxide degradation is a result of the oxide and interfacial traps generation during FN (Fowler-Nordheim) stress [2]. Because the bias temperature stress causes an increase of both interfacial-traps and fixed oxide charge could be a factor, witch can degrade device reliability during the program and erase operation. However, few studies on NBTI/PBTI have been conducted on improving the reliability of MONOS devices. In this work, we investigate the effect of post-annealing gas on bias temperature instability (BTI), such as the flat band voltage shift, ${\Delta}V_{FB}$, the interfacial trap density shift, ${\Delta}I_G$ retention characteristics, and the gate leakage current characteristics of MANOS capacitors. MANOS samples annealed at $950^{\circ}C$ for 30 s by a rapid thermal process were treated via additional annealing in a furnace, using annealing gases $N_2$ and $N_2-H_2$ (2 % hydrogen and 98 % nitrogen mixture gases) at $450^{\circ}C$ for 30 min. MANOS samples annealed in $N_2-H_2$ ambient had the lowest flat band voltage shift, ${\Delta}V_{FB}$ = 1.09/0.63 V at the program/erase state, and the good retention characteristics, 123/84 mV/decade at the program/erase state more than the sample annealed at $N_2$ ambient.

• Journal of the Korean Society of Fisheries and Ocean Technology
• /
• v.26 no.3
• /
• pp.265-274
• /
• 1990

The temperature inversions were studied on the basis of Digital Memory Bathythermography(DBT) data collected by training ship, Pusan 402, of the National Fisheries University of Pusan in August 23~25, 1986 and Fisheries Reserach and Development Agency of Korea in August, 1986, The results were as follows; Among the 67 stations of studied area, occurrence frequency of temperature inversion was 58.20%, And the frequency of onefold occurrence of temperature inversion at its profile of each station was 13.42%. of twofold occurrence was 20.80%, and of threefold occurrence was 23.88%. In the studied area, the temperature inversion usually occurred below the 40m depth and its layers also located below the thermocline. The temperature range of its inversion was from 14$^{\circ}C$ to 16$^{\circ}C$. The temperature inversion in the study area was oaused by the interaction between Tsushima Warm Current and Korea Coastal Waters.

• Journal of Broadcast Engineering
• /
• v.25 no.4
• /
• pp.545-559
• /
• 2020

In this paper, we propose a content creation framework that enables users without programming experience to easily create interactive media content that responds to user gestures. In the proposed framework, users define the gestures they use and the media effects that respond to them by numbers, and link them in a text-based configuration file. In the proposed framework, the interactive media content that responds to the user's gesture is linked with the dynamic projection mapping module to track the user's location and project the media effects onto the user. To reduce the processing speed and memory burden of the gesture recognition, the user's movement is expressed as a gray scale motion history image. We designed a convolutional neural network model for gesture recognition using motion history images as input data. The number of network layers and hyperparameters of the convolutional neural network model were determined through experiments that recognize five gestures, and applied to the proposed framework. In the gesture recognition experiment, we obtained a recognition accuracy of 97.96% and a processing speed of 12.04 FPS. In the experiment connected with the three media effects, we confirmed that the intended media effect was appropriately displayed in real-time according to the user's gesture.

• Journal of the Korean Society of Radiology
• /
• v.16 no.4
• /
• pp.453-462
• /
• 2022

In this paper, the artificial intelligence (AI) technology used in the medical image analysis field was analyzed through a literature review. Literature searches were conducted on PubMed, ResearchGate, Google and Cochrane Review using the key word. Through literature search, 114 abstracts were searched, and 98 abstracts were reviewed, excluding 16 duplicates. In the reviewed literature, AI is applied in classification, localization, disease detection, disease segmentation, and fit degree of registration images. In machine learning (ML), prior feature extraction and inputting the extracted feature values into the neural network have disappeared. Instead, it appears that the neural network is changing to a deep learning (DL) method with multiple hidden layers. The reason is thought to be that feature extraction is processed in the DL process due to the increase in the amount of memory of the computer, the improvement of the calculation speed, and the construction of big data. In order to apply the analysis of medical images using AI to medical care, the role of physicians is important. Physicians must be able to interpret and analyze the predictions of AI algorithms. Additional medical education and professional development for existing physicians is needed to understand AI. Also, it seems that a revised curriculum for learners in medical school is needed.

• Journal of the Korea Society of Computer and Information
• /
• v.17 no.6
• /
• pp.1-11
• /
• 2012

As the process technology scales down, multi-core processors cause serious problems such as increased interconnection delay, high power consumption and thermal problems. To solve the problems in 2D multi-core processors, researchers have focused on the 3D multi-core processor architecture. Compared to the 2D multi-core processor, the 3D multi-core processor decreases interconnection delay by reducing wire length significantly, since each core on different layers is connected using vertical through-silicon via(TSV). However, the power density in the 3D multi-core processor is increased dramatically compared to that in the 2D multi-core processor, because multiple cores are stacked vertically. Unfortunately, increased power density causes thermal problems, resulting in high cooling cost, negative impact on the reliability. Therefore, temperature should be considered together with performance in designing 3D multi-core processors. In this work, we analyze the temperature of the cache in quad-core processors varying cache organization. Then, we propose the low-temperature cache organization to overcome the thermal problems. Our evaluation shows that peak temperature of the instruction cache is lower than threshold. The peak temperature of the data cache is higher than threshold when the cache is composed of many ways. According to the results, our proposed cache organization not only efficiently reduces the peak temperature but also reduces the performance degradation for 3D quad-core processors.

• Journal of the Korean Magnetics Society
• /
• v.10 no.5
• /
• pp.210-219
• /
• 2000

The magnetoresistance characteristics of FeN/Co/Cu/Co and FeN/Co/Cu/Co/Cu/Co/FeN multilayers using ferromagnetic iron-nitrides (FeN) has been studied. The microstructure of FeN film is the mixed ${\alpha}$-Fe and $\varepsilon$-Fe$_3$N phase on the condition that the flow rate of N$_2$ gas is over 0.4 sccm. The magnetoresistance effect is observed because of shape magnetic anisotropy induced by needle-shaped $\varepsilon$-Fe$_3$N phase. This magnetoresistance effect changes, because the degree that the shape magnetic anisotropy adheres to the adjacent Co pinned layer is varied according to the flow rate of N$_2$ gas and the thickness of FeN film. The best magnetoresistance effect is obtained on the condition that the thickness of Co free layer is 70 ${\AA}$ and the maximum MR ratio(%) value of 3.2% shows in the FeN(250 ${\AA}$)/Co(70 ${\AA}$)/Cu(25 ${\AA}$)/Co(70 ${\AA}$)/Cu(25 ${\AA}$)/Co(70 ${\AA}$)/FeN(250 ${\AA}$) mutilayer film which is fabricated at the N, gas flow rate of 0.5 sccm and the FeN film thickness of 250 ${\AA}$. Four steps are observed in the magnetoresistance curve owing to this difference of coercive force, because respective magnetic layers in the multilayer possess different coercive forces. These effects observed in these mutilayer films can be expected to application to the memory device the same MRAM as can carry out simultaneously four signals.

• Journal of the Institute of Electronics and Information Engineers
• /
• v.51 no.10
• /
• pp.96-106
• /
• 2014

Previous academic research on FPGA tools has relied on simple imaginary models for the targeting architecture. As the first step to overcome such restriction, the issues on analytic placement and legalization which are applied to commercial FPGAs have been brought up, and several techniques to remedy them are presented, and evaluated. First of all, the center of gravity of the placed cells may be far displaced from the center of the chip during analytic placement. A function is proposed to be added to the objective function for minimizing this displacement. And then, the density map is expanded into multiple layers to accurately calculate the density distribution for each of the cell types. Early fixation is also proposed for the memory blocks which can be placed at limited sites in small numbers. Since two flip-flops share control pins in a slice, a compatibility constraint is introduced during legalization. Pre-packing compatible flip-flops is proposed as a proactive step. The proposed techniques are implemented on the K-FPGA fabric evaluation framework in which commercial architectures can be precisely modeled, and modified for enhancement, and validated on twelve industrial strength examples. The placement results show that the proposed techniques have reduced the wire length by 22%, and the slice usage by 5% on average. This research is expected to be a development basis of the optimization CAD tools for new as well as the state-of-the-art FPGA architectures.

• Journal of the Computational Structural Engineering Institute of Korea
• /
• v.30 no.5
• /
• pp.389-396
• /
• 2017

Peridynamics has been widely used in the dynamic fracture analysis of brittle materials. Recently, various crack patterns(compact region, floret, Hertz-type crack, etc.) of multilayered glass structures in experiments(Bless et al. 2010) were implemented with a bond-based peridynamic simulation(Bobaru et al.. 2012). The actual glass layers are bound with thin elastic interlayer material while the interlayer is missing from the peridynamic model used in the previous numerical study. In this study, the peridynamic interlayer modeling for the multilayered structures is proposed. It requires enormous computational time and memory to explicitly model very thin interlayer materials. Instead of explicit modeling, fictitious peridynamic particles are introduced for modeling interlayer materials. The computational efficiency and accuracy of the proposed peridynamic interlayer model are verified through numerical tests. Furthermore, preventing penetration scheme based on short-range interaction force is employed for the multilayered structure under compression and verified through parametric tests.

• Cartoon and Animation Studies
• /
• s.39
• /
• pp.111-132
• /
• 2015

Although Rotoscoping technique has been used for a long time to mimic the natural and smooth motion since the early 20th century, its artistic value was devalued as tricks because it traces the already recorded images. But the fact that the rotoscoping images can cross the boundaries between animation and live action in an infinite integral freedom in the digital era became rather expansive new aesthetic possibilities of representation of the reality. In addition, Jonas Odell's animations such as (2010), (2008), (2006) are good example to prove that the rotoscoping images also can serve as means to enhance its narrative. This study is to analyze how rotoscoping images act as a unique role in relation with the narrative based on the said person's real stories and realistic images. I argue that his animated films constantly contain these three characteristics -Images to mediate Auditory sensitivity as a record of inner metaphysical reality, anonymous images to represent a specific existential character, and images that act as physical representation that holds the physical space/time and related memory. This reveals that rotoscoping images in this digital era went beyond reproduction for natural movements or special type of style. It rather suggests new layers of experience, and acquires new value in animation. I hope that this study could serve as a foundation to rediscover and re-position the value of rotoscoping images as well as good opportunity to introduce very talented swedish animation artist who already received global attention with his unique philosophical and aesthetic style.