• 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.

• Journal of IKEEE
• /
• v.18 no.4
• /
• pp.593-602
• /
• 2014

One of the purposes of an artificial neural netowrk(ANNet) is to implement the largest number of functions as possible with the smallest number of nodes and layers. This paper presents a CoreNet which has a multi-leveled input value and a multi-leveled output value with a 2-layered ANNet, which is the basic structure of an ANNet. I have suggested an equation for calculating the capacity of the CoreNet, which has a p-leveled input and a q-leveled output, as $a_{p,q}={\frac{1}{2}}p(p-1)q^2-{\frac{1}{2}}(p-2)(3p-1)q+(p-1)(p-2)$. I've applied this CoreNet into the simulation model 1(5)-1(6), which has 5 levels of an input and 6 levels of an output with no hidden layers. The simulation result of this model gives, the maximum 219 convergences for the number of implementable functions using the cot(${\sqrt{x}}$) input leveling method. I have also shown that, the 27 functions are implementable by the calculation of weight values(w, ${\theta}$) with the multi-threshold lines in the weight space, which are diverged in the simulation results. Therefore the 246 functions are implementable in the 1(5)-1(6) model, and this coincides with the value from the above eqution $a_{5,6}(=246)$. I also show the implementable function numbering method in the weight space.

• Korean Journal of Construction Engineering and Management
• /
• v.23 no.3
• /
• pp.45-55
• /
• 2022

BIM models allow building spaces to be instantiated and recognized as unique objects independently of model elements. These instantiated spaces provide the required semantics that can be leveraged for building code checking, energy analysis, and evacuation route analysis. However, theses spaces or rooms need to be designated manually, which in practice, lead to errors and omissions. Thus, most BIM models today does not guarantee the semantic integrity of space designations, limiting their potential applicability. Recent studies have explored ways to automate space allocation in BIM models using artificial intelligence algorithms, but they are limited in their scope and relatively low classification accuracy. This study explored the use of Graph Convolutional Networks, an algorithm exclusively tailored for graph data structures. The goal was to utilize not only geometry information but also the semantic relational data between spaces and elements in the BIM model. Results of the study confirmed that the accuracy was improved by about 8% compared to algorithms that only used geometric distinctions of the individual spaces.

• Journal of the Korea Convergence Society
• /
• v.13 no.3
• /
• pp.77-89
• /
• 2022

It has been proven through many previous researches that the pretrained language model with a large corpus helps improve performance in various natural language processing tasks. However, there is a limit to building a large-capacity corpus for training in a language environment where resources are scarce. Using the Cross-lingual Post-Training (XPT) method, we analyze the method's efficiency in Korean, which is a low resource language. XPT selectively reuses the English pretrained language model parameters, which is a high resource and uses an adaptation layer to learn the relationship between the two languages. This confirmed that only a small amount of the target language dataset in the relationship extraction shows better performance than the target pretrained language model. In addition, we analyze the characteristics of each model on the Korean language model and the Korean multilingual model disclosed by domestic and foreign researchers and companies.

• Journal of the Korean Geotechnical Society
• /
• v.39 no.9
• /
• pp.13-24
• /
• 2023

To investigate the stability of temporary retaining walls during excavation, it is essential to develop reverse analysis technologies capable of precisely evaluating the properties of the ground and a learning model that can assess stability by analyzing real-time data. In this study, we targeted excavation sites where the C.I.P method was applied. We developed a Deep Neural Network (DNN) model capable of evaluating the stability of the retaining wall, and estimated the physical properties of the ground being excavated using a Differential Evolution Algorithm. We performed reverse analysis on a model composed of a two-layer ground for the applicability analysis of the Differential Evolution Algorithm. The results from this analysis allowed us to predict the properties of the ground, such as the elastic modulus, cohesion, and internal friction angle, with an accuracy of 97%. We analyzed 30,000 cases to construct the training data for the DNN model. We proposed stability evaluation grades for each assessment factor, including anchor axial force, uneven subsidence, wall displacement, and structural stability of the wall, and trained the data based on these factors. The application analysis of the trained DNN model showed that the model could predict the stability of the retaining wall with an average accuracy of over 94%, considering factors such as the axial force of the anchor, uneven subsidence, displacement of the wall, and structural stability of the wall.

• Journal of Intelligence and Information Systems
• /
• v.27 no.4
• /
• pp.1-22
• /
• 2021

Aspect Based Sentiment Analysis (ABSA), which analyzes sentiment based on aspects that appear in the text, is drawing attention because it can be used in various business industries. ABSA is a study that analyzes sentiment by aspects for multiple aspects that a text has. It is being studied in various forms depending on the purpose, such as analyzing all targets or just aspects and sentiments. Here, the aspect refers to the property of a target, and the target refers to the text that causes the sentiment. For example, for restaurant reviews, you could set the aspect into food taste, food price, quality of service, mood of the restaurant, etc. Also, if there is a review that says, "The pasta was delicious, but the salad was not," the words "steak" and "salad," which are directly mentioned in the sentence, become the "target." So far, in ABSA, most studies have analyzed sentiment only based on aspects or targets. However, even with the same aspects or targets, sentiment analysis may be inaccurate. Instances would be when aspects or sentiment are divided or when sentiment exists without a target. For example, sentences like, "Pizza and the salad were good, but the steak was disappointing." Although the aspect of this sentence is limited to "food," conflicting sentiments coexist. In addition, in the case of sentences such as "Shrimp was delicious, but the price was extravagant," although the target here is "shrimp," there are opposite sentiments coexisting that are dependent on the aspect. Finally, in sentences like "The food arrived too late and is cold now." there is no target (NULL), but it transmits a negative sentiment toward the aspect "service." Like this, failure to consider both aspects and targets - when sentiment or aspect is divided or when sentiment exists without a target - creates a dual dependency problem. To address this problem, this research analyzes sentiment by considering both aspects and targets (Target-Aspect-Sentiment Detection, hereby TASD). This study detected the limitations of existing research in the field of TASD: local contexts are not fully captured, and the number of epochs and batch size dramatically lowers the F1-score. The current model excels in spotting overall context and relations between each word. However, it struggles with phrases in the local context and is relatively slow when learning. Therefore, this study tries to improve the model's performance. To achieve the objective of this research, we additionally used auxiliary loss in aspect-sentiment classification by constructing CNN(Convolutional Neural Network) layers parallel to existing models. If existing models have analyzed aspect-sentiment through BERT encoding, Pooler, and Linear layers, this research added CNN layer-adaptive average pooling to existing models, and learning was progressed by adding additional loss values for aspect-sentiment to existing loss. In other words, when learning, the auxiliary loss, computed through CNN layers, allowed the local context to be captured more fitted. After learning, the model is designed to do aspect-sentiment analysis through the existing method. To evaluate the performance of this model, two datasets, SemEval-2015 task 12 and SemEval-2016 task 5, were used and the f1-score increased compared to the existing models. When the batch was 8 and epoch was 5, the difference was largest between the F1-score of existing models and this study with 29 and 45, respectively. Even when batch and epoch were adjusted, the F1-scores were higher than the existing models. It can be said that even when the batch and epoch numbers were small, they can be learned effectively compared to the existing models. Therefore, it can be useful in situations where resources are limited. Through this study, aspect-based sentiments can be more accurately analyzed. Through various uses in business, such as development or establishing marketing strategies, both consumers and sellers will be able to make efficient decisions. In addition, it is believed that the model can be fully learned and utilized by small businesses, those that do not have much data, given that they use a pre-training model and recorded a relatively high F1-score even with limited resources.

• Korean Journal of Remote Sensing
• /
• v.38 no.6_1
• /
• pp.1463-1478
• /
• 2022

Maritime transport accounts for 99.7% of the exports and imports of the Republic of Korea; therefore, developing a vessel monitoring system for efficient operation is of significant interest. Several studies have focused on tracking and monitoring vessel movements based on automatic identification system (AIS) data; however, ships without AIS have limited monitoring and tracking ability. High-resolution optical satellite images can provide the missing layer of information in AIS-based monitoring systems because they can identify non-AIS vessels and small ships over a wide range. Therefore, it is necessary to investigate vessel monitoring and small vessel classification systems using high-resolution optical satellite images. This study examined the possibility of developing ship monitoring systems using Compact Advanced Satellite 500-1 (CAS500-1) satellite images by first training a deep learning model using satellite image data and then performing detection in other images. To determine the effectiveness of the proposed method, the learning data was acquired from ships in the Yellow Sea and its major ports, and the detection model was established using the You Only Look Once (YOLO) algorithm. The ship detection performance was evaluated for a domestic and an international port. The results obtained using the detection model in ships in the anchorage and berth areas were compared with the ship classification information obtained using AIS, and an accuracy of 85.5% and 70% was achieved using domestic and international classification models, respectively. The results indicate that high-resolution satellite images can be used in mooring ships for vessel monitoring. The developed approach can potentially be used in vessel tracking and monitoring systems at major ports around the world if the accuracy of the detection model is improved through continuous learning data construction.

• 대한공업교육학회지
• /
• v.44 no.2
• /
• pp.104-122
• /
• 2019

In the era of the fourth industrial revolution, maker education is drawing attention as a method of student-led education. At a time when interest in maker education is also growing in technology education, figuring out what stage of concern(SoC) a middle school technology teacher is critical to effective implementation. This study analyzed SoC in maker education by layer sampling among 400 middle school technology teachers using Concerns-based adoption model. SoC was then obtained by measuring the origin using the SoCQ and then presenting it as a SOCQ profile. Gender, training experience with two lower variables were analyzed using t verification, working cities, teaching experience with more than three lower variables were analyzed using one-way ANOVA. Studies showed that SoC in maker education of middle school technology teachers showed the most similar characteristics to that of non-users. The difference in concern depending on gender was that male teachers were more concerned in maker education than female teachers. The difference in concern depending on the working city was that teachers working in the township were more concerned in the maker education than teachers working in the large city, and the difference in concern depending on the teaching career was higher among teachers with middle experience than those with low and high experience. There was also a higher stage of concern in maker education than in teachers without training experience. Therefore, it is necessary to provide middle school technology teachers with an introduction to the maker education and various information, teaching, learning and evaluation data to enhance overall concern and to support the use and evaluation of the maker education in the classroom by providing various teacher training and consulting on the maker education in the future. Further, through further study, we should conduct study that analyzes both Stage of Concern, Level of Use and Innovation Configuration, to put in the effort for effective settlement of maker education.

• Proceedings of the Optical Society of Korea Conference
• /
• 2003.07a
• /
• pp.60-61
• /
• 2003

A new approach to reduce the computation time of genetic algorithm (GA) for making binary phase holograms is described. Synthesized holograms having diffraction efficiency of 75.8% and uniformity of 5.8% are proven in computer simulation and experimentally demonstrated. Recently, computer-generated holograms (CGHs) having high diffraction efficiency and flexibility of design have been widely developed in many applications such as optical information processing, optical computing, optical interconnection, etc. Among proposed optimization methods, GA has become popular due to its capability of reaching nearly global. However, there exits a drawback to consider when we use the genetic algorithm. It is the large amount of computation time to construct desired holograms. One of the major reasons that the GA' s operation may be time intensive results from the expense of computing the cost function that must Fourier transform the parameters encoded on the hologram into the fitness value. In trying to remedy this drawback, Artificial Neural Network (ANN) has been put forward, allowing CGHs to be created easily and quickly (1), but the quality of reconstructed images is not high enough to use in applications of high preciseness. For that, we are in attempt to find a new approach of combiningthe good properties and performance of both the GA and ANN to make CGHs of high diffraction efficiency in a short time. The optimization of CGH using the genetic algorithm is merely a process of iteration, including selection, crossover, and mutation operators [2]. It is worth noting that the evaluation of the cost function with the aim of selecting better holograms plays an important role in the implementation of the GA. However, this evaluation process wastes much time for Fourier transforming the encoded parameters on the hologram into the value to be solved. Depending on the speed of computer, this process can even last up to ten minutes. It will be more effective if instead of merely generating random holograms in the initial process, a set of approximately desired holograms is employed. By doing so, the initial population will contain less trial holograms equivalent to the reduction of the computation time of GA's. Accordingly, a hybrid algorithm that utilizes a trained neural network to initiate the GA's procedure is proposed. Consequently, the initial population contains less random holograms and is compensated by approximately desired holograms. Figure 1 is the flowchart of the hybrid algorithm in comparison with the classical GA. The procedure of synthesizing a hologram on computer is divided into two steps. First the simulation of holograms based on ANN method [1] to acquire approximately desired holograms is carried. With a teaching data set of 9 characters obtained from the classical GA, the number of layer is 3, the number of hidden node is 100, learning rate is 0.3, and momentum is 0.5, the artificial neural network trained enables us to attain the approximately desired holograms, which are fairly good agreement with what we suggested in the theory. The second step, effect of several parameters on the operation of the hybrid algorithm is investigated. In principle, the operation of the hybrid algorithm and GA are the same except the modification of the initial step. Hence, the verified results in Ref [2] of the parameters such as the probability of crossover and mutation, the tournament size, and the crossover block size are remained unchanged, beside of the reduced population size. The reconstructed image of 76.4% diffraction efficiency and 5.4% uniformity is achieved when the population size is 30, the iteration number is 2000, the probability of crossover is 0.75, and the probability of mutation is 0.001. A comparison between the hybrid algorithm and GA in term of diffraction efficiency and computation time is also evaluated as shown in Fig. 2. With a 66.7% reduction in computation time and a 2% increase in diffraction efficiency compared to the GA method, the hybrid algorithm demonstrates its efficient performance. In the optical experiment, the phase holograms were displayed on a programmable phase modulator (model XGA). Figures 3 are pictures of diffracted patterns of the letter "0" from the holograms generated using the hybrid algorithm. Diffraction efficiency of 75.8% and uniformity of 5.8% are measured. We see that the simulation and experiment results are fairly good agreement with each other. In this paper, Genetic Algorithm and Neural Network have been successfully combined in designing CGHs. This method gives a significant reduction in computation time compared to the GA method while still allowing holograms of high diffraction efficiency and uniformity to be achieved. This work was supported by No.mOl-2001-000-00324-0 (2002)) from the Korea Science & Engineering Foundation.

• Journal of the Korean earth science society
• /
• v.42 no.3
• /
• pp.247-263
• /
• 2021

Climate change has been accelerating in coastal waters recently; therefore, the importance of coastal environmental monitoring is also increasing. Chlorophyll-a concentration, an important marine variable, in the surface layer of the global ocean has been retrieved for decades through various ocean color satellites and utilized in various research fields. However, the commonly used chlorophyll-a concentration algorithm is only suitable for application in clear water and cannot be applied to turbid waters because significant errors are caused by differences in their distinct components and optical properties. In addition, designing a standard algorithm for coastal waters is difficult because of differences in various optical characteristics depending on the coastal area. To overcome this problem, various algorithms have been developed and used considering the components and the variations in the optical properties of coastal waters with high turbidity. Chlorophyll-a concentration retrieval algorithms can be categorized into empirical algorithms, semi-analytic algorithms, and machine learning algorithms. These algorithms mainly use the blue-green band ratio based on the reflective spectrum of sea water as the basic form. In constrast, algorithms developed for turbid water utilizes the green-red band ratio, the red-near-infrared band ratio, and the inherent optical properties to compensate for the effect of dissolved organisms and suspended sediments in coastal area. Reliable retrieval of satellite chlorophyll-a concentration from turbid waters is essential for monitoring the coastal environment and understanding changes in the marine ecosystem. Therefore, this study summarizes the pre-existing algorithms that have been utilized for monitoring turbid Case 2 water and presents the problems associated with the mornitoring and study of seas around the Korean Peninsula. We also summarize the prospective for future ocean color satellites, which can yield more accurate and diverse results regarding the ecological environment with the development of multi-spectral and hyperspectral sensors.