• Journal of the Korea Society of Computer and Information
• /
• v.26 no.9
• /
• pp.57-64
• /
• 2021

In this paper, we propose a device that can control the primary control surface of an aircraft by recognizing speech commands. The speech command consists of 19 commands, and a learning model is constructed based on a total of 2,500 datasets. The training model is composed of a CNN model using the Sequential library of the TensorFlow-based Keras model, and the speech file used for training uses the MFCC algorithm to extract features. The learning model consists of two convolution layers for feature recognition and Fully Connected Layer for classification consists of two dense layers. The accuracy of the validation dataset was 98.4%, and the performance evaluation of the test dataset showed an accuracy of 97.6%. In addition, it was confirmed that the operation was performed normally by designing and implementing a Raspberry Pi-based control device. In the future, it can be used as a virtual training environment in the field of voice recognition automatic flight and aviation maintenance.

• KIPS Transactions on Software and Data Engineering
• /
• v.10 no.2
• /
• pp.65-72
• /
• 2021

Recently, attacks on the network environment have been rapidly escalating and intelligent. Thus, the signature-based network intrusion detection system is becoming clear about its limitations. To solve these problems, research on machine learning-based intrusion detection systems is being conducted in many ways, but two problems are encountered to use machine learning for intrusion detection. The first is to find important features associated with learning for real-time detection, and the second is the imbalance of data used in learning. This problem is fatal because the performance of machine learning algorithms is data-dependent. In this paper, we propose the HSF-DNN, a network intrusion detection model based on a deep neural network to solve the problems presented above. The proposed HFS-DNN was learned through the NSL-KDD data set and performs performance comparisons with existing classification models. Experiments have confirmed that the proposed Hybrid Feature Selection algorithm does not degrade performance, and in an experiment between learning models that solved the imbalance problem, the model proposed in this paper showed the best performance.

• The Journal of Korea Institute of Information, Electronics, and Communication Technology
• /
• v.14 no.3
• /
• pp.223-230
• /
• 2021

Currently, self-driving vehicles are on the verge of commercialization after testing. However, even though autonomous vehicles have not been fully commercialized, 81 accidents have occurred, and the driving method of vehicles to avoid accidents relies heavily on LiDAR. In order for the currently commercialized 3-level autonomous vehicle to develop into a 4-level autonomous vehicle, more information must be collected than previously collected information. Therefore, this paper proposes a Driving Situation Judgment System (DSJS) that accurately calculates the crisis situation the vehicle is in by useing the roughness of the road and the state of the passengers of surrounding vehicles including road information and weather information collected from existing autonomous vehicles. As a result of DSJS's PDM experiment, PDM was able to classify passengers 15.52% more accurately on average than the existing vehicle's passenger recognition system. This study can be a basic research to achieve the 4th level autonomous vehicle by collecting more various types than the data collected by the existing 3rd level autonomous vehicle.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.24 no.1
• /
• pp.57-64
• /
• 2020

Recently, due to the proliferation of IoT sensors, the development of big data and artificial intelligence, time series prediction research on fine dust pollution is actively conducted. However, because the data representing fine dust contamination changes rapidly, traditional time series prediction methods do not provide a level of accuracy that can be used in the field. In this paper, we propose a method that reflects the classification results of environmental conditions through CNN when predicting micro dust contamination using LSTM. Although LSTM and CNN are independent, they are integrated into one network through the interface, so this method is easier to understand than the application LSTM. In the verification experiments of the proposed method using Beijing PM2.5 data, the prediction accuracy and predictive power for the timing of change were consistently improved in various experimental cases.

• KIPS Transactions on Software and Data Engineering
• /
• v.10 no.6
• /
• pp.223-234
• /
• 2021

Recent advances in the 4th Industrial Revolution have accelerated the change of the shopping behavior from offline to online. Search queries show customers' information needs most intensively in online shopping. However, there are not many search query research in the field of search, and most of the prior research in the field of search query research has been studied on a limited topic and data-based basis based on researchers' qualitative judgment. To this end, this study defines the type of search query with data-based quantitative methodology by applying machine learning to search research query field to define the 15 topics of search query by conducting topic modeling based on search query and clicked document information. Furthermore, we present a new classification system of new search query types representing searching behavior characteristics by extracting key variables through principal component analysis and analyzing. The results of this study are expected to contribute to the establishment of effective search services and the development of search systems.

• The Journal of Information Systems
• /
• v.30 no.4
• /
• pp.277-301
• /
• 2021

Purpose The purpose of this study is to analyze the sentiment responses of the general public to non-face-to-face work using text mining methodology. As the number of non-face-to-face complaints is increasing over time, it is difficult to review and analyze in traditional methods such as surveys, and there is a limit to reflect real-time issues. Approach This study has proposed a method of the research model, first by collecting and cleansing the data related to non-face-to-face work among tweets posted on Twitter. Second, topics and keywords are extracted from tweets using LDA(Latent Dirichlet Allocation), a topic modeling technique, and changes for each section are analyzed through DTM(Dynamic Topic Modeling). Third, the complaints of non-face-to-face work are analyzed through the classification of positive and negative polarity in the COVID-19 section. Findings As a result of analyzing 1.54 million tweets related to non-face-to-face work, the number of IDs using non-face-to-face work-related words increased 7.2 times and the number of tweets increased 4.8 times after COVID-19. The top frequently used words related to non-face-to-face work appeared in the order of remote jobs, cybersecurity, technical jobs, productivity, and software. The words that have increased after the COVID-19 were concerned about lockdown and dismissal, and business transformation and also mentioned as to secure business continuity and virtual workplace. New Normal was newly mentioned as a new standard. Negative opinions found to be increased in the early stages of COVID-19 from 34% to 43%, and then stabilized again to 36% through non-face-to-face work sentiment analysis. The complaints were, policies such as strengthening cybersecurity, activating communication to improve work productivity, and diversifying work spaces.

• Journal of the Institute of Convergence Signal Processing
• /
• v.22 no.4
• /
• pp.206-212
• /
• 2021

Society is developing more and more, and communication methods are developing in many ways. However, developed communication is a way for the non-disabled and has no effect on the deaf. Therefore, in this paper, a CNN-based sign language translation program is designed and implemented to help deaf people communicate. Sign language translation programs translate sign language images entered through WebCam according to meaning based on data. The sign language translation program uses 24,000 pieces of Korean vowel data produced directly and conducts U-Net segmentation to train effective classification models. In the implemented sign language translation program, 'ㅋ' showed the best performance among all sign language data with 97% accuracy and 99% F1-Score, while 'ㅣ' showed the highest performance among vowel data with 94% accuracy and 95.5% F1-Score.

• Journal of the Society of Naval Architects of Korea
• /
• v.59 no.4
• /
• pp.192-199
• /
• 2022

With the advent of autonomous ships, it is emerging as one of the very important issues not only to operate with a minimum crew or unmanned ships, but also to secure the safety of ships to prevent marine accidents. On-site inspection of the hull is mainly performed by the inspector's visual inspection, and video information is recorded using a small camera if necessary. However, due to the shortage of inspection personnel, time and space constraints, and the pandemic situation, the necessity of introducing an automated inspection system using artificial intelligence and remote inspection is becoming more important. Furthermore, research on hardware and software that enables the automated inspection system to operate normally even under the harsh environmental conditions of a ship is absolutely necessary. For automated inspection systems, it is important to review artificial intelligence technologies and equipment that can perform a variety of hull failure detection and classification. To address this, it is important to classify the hull failure. Based on various guidelines and expert opinions, we divided them into 6 types(Crack, Corrosion, Pitting, Deformation, Indent, Others). It was decided to apply object detection technology to cracks of hull failure. After that, YOLOv5 was decided as an artificial intelligence model suitable for survey and a common hull crack dataset was trained. Based on the performance results, it aims to present the possibility of applying artificial intelligence in the field by determining and testing the equipment required for survey.

• Journal of the Korean GEO-environmental Society
• /
• v.23 no.12
• /
• pp.17-23
• /
• 2022

This paper analyze the precast panel retaining wall's safety factor changes based on the numerical analysis using PLAXIS 2D software. Numerical analysis conditions include construction method, nail and panel fixing method, backfill material compaction conditions, rainfall conditions. The classification according to the construction method of the precast panel retaining wall includes the top-down and bottom-up methods. The difference between the top-down and bottom-up methods is the presence or absence backfill material and the ground excavation method. The top-down method involves vertically excavating the ground and attaching the panel using mortar, but in the bottom-up method, the ground is vertically excavated and harden the backfill material. As a result of numerical analysis, the top-down method secured a higher safety factor in all cases except the rainfall conditions.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.27 no.1
• /
• pp.69-75
• /
• 2023

Small communication devices used in the Internet of Things are vulnerable to various hacking because they do not apply advanced encryption techniques due to their low memory capacity or slow computation speed. In order to increase the authentication reliability of small-sized transmitters operating in 433MHz band, we introduce an RF fingerprint and adopt a convolutional neural network (CNN) as a classification algorithm. The preamble signal transmitted by each transmitter are extracted and collected using software-defined-radio to constitute a training data set, which is used for training the CNN. We tested identification of 20 transmitters in four different scenarios and obtained high identification accuracy. In particular, the accuracy of 95.8% and 92.6% was obtained, respectively in the scenario where the test was performed at a location different from the transmitter's location at the time of collecting training data, and in the scenario where the transmitter moves at walking speed.