Korean | English

pISSN : 1975-7700 / eISSN : 2734-0570

2020 KCI Impact Factor : 0.42
Home > Explore Content > All Issues > Article List

2018, Vol.13, No.4

  • 1.

    A Study of Heart Disease Prediction Using Multilayer Perceptron based on Deep Learning

    Kim Chul Jin | Kim JiSeong | 2018, 13(4) | pp.411~419 | number of Cited : 0
    Abstract PDF
    Deep Learning, an advanced technology for machine learning, is currently being applied to various industrial fields and is being actively applied and developed in the fields of finance, e-commerce, and Internet of Thing, which require prediction. In particular, this study will be applied to the prediction of heart disease in the medical field, and it will provide the advantage of early detection of heart disease or saving medical treatment cost such as angiography for non-heart disease patient. To apply the Multilayer Perceptron machine learning technique for Deep Learning, 2 Hidden Layers and 10 Perceptons can be constructed to improve learning accuracy. By using Back Propagation during learning, it is reversed from the Output Layer to the Hidden Layer so as to reduce the error. The error is reduced by using ReLU or Sigmoid function, and the generated prediction model optimizes the model through the Adam optimization function. In a case study, heart disease data is learned using heart disease diagnosis and presence or absence of heart disease provided by the Machine Learning & Artificial Intelligence System Center at the University of California, Irvine. The Deep Learning module is developed using a Python-based TensorFlow and validated using randomly extracted data samples from heart disease data after learning.
  • 2.

    Interactive Data Visualization Based Realtime Monitoring and Fault Detection System

    Seung-Taek Ryoo | 2018, 13(4) | pp.421~428 | number of Cited : 0
    Abstract PDF
    Effective representation of various types of data from various types of equipment is not a trivial task. The more data items collected for sophisticated data analysis, the more trouble the system operator has to identify important items. In order to solve these problems, it is necessary to develop a real-time interactive data visualization system that can be efficiently observing large-volume data at a glance and confirming fault detection in advance. In this paper, we propose a real - time observation and fault detection system using interactive data visualization method to efficiently manage consecutively generated raw data. The proposed system consists of data collection, data visualization, data fault detection and alarm steps. In the data collection step, the real-time generated device data using the data collection device is stored, and the collected data is massaged and generalized through data preprocessing process. The normalized data is subjected to data analysis, visualization mapping, and rendering in the data visualization step. To visualize and render the defined data primitives, we use heat map, radar chart, and stream graph. In addition, statistical process control (SPC) rule-based fault detection and alarm methods are used to detect data errors. The proposed method can be widely used in various fields such as image generation, data fusion, fault detection and machine learning.
  • 3.

    Development of Situation Awareness System for Emergency Situation of Single-person Households

    Kim Hye Ri | Lee Sang Joon | 2018, 13(4) | pp.429~439 | number of Cited : 0
    Abstract PDF
    Social issues such as isolation and lonely death of single-person households who are socially marginalized and disconnected are occurred. It is necessary to develop a situational awareness system to prevent the emergency situation of single-person households. In previous researches with contact type bio-signal measurement system, there are limitations of real-time monitoring due to contact failure or continuous measurement problem. Also, in the conventional non-contact type bio-signal system, the method using camera has caused errors due to external factors such as light or glasses, and there are limitations such as expensive bio-signal measurement sensors, complicated installation, and space utilization problems. In this paper, we have developed a system that can overcome these drawbacks. We designed the UWB Radar signal processing algorithm, subject position tracking algorithm, and breathing measurement algorithm. We have used the TOA(Time of Arrival) technique for position measurement, Kalman filter and background difference algorithm for noise reduction in signal processing with IR-UWB Radar. The proposed system was tested in a 4M * 4M room, similar to the real one, with emergency scenario of one person. The position of the subject was accurately tracked and the position error within 20cm was obtained. The system developed in this paper can be used by government and corporations to carry out social welfare and emergency rescue services by using online circuits through G2B(Goverment to Business).
  • 4.

    A Development Architecture Research of Intelligent Object Recognition and Voice Service for the Visually Impaired

    Kim Chul Jin | Myung Soo Park | Min Hwan Kim | 2018, 13(4) | pp.441~450 | number of Cited : 1
    Abstract PDF
    The development of information technology has made daily life convenient and enjoys cultural benefits through various information technology services. However, due to the limited accessibility of information technology, disabled people do not receive various information technology services. Therefore, in this paper, we propose object recognition and voice service architecture to improve accessibility of information technology service for the disabled. In particular, it enables the visually impaired to receive recognition services and voice services for objects. The visually impaired person recognizes the desired objects using the mobile device, and the recognized object information is provided by voice so that the visually impaired person can recognize the object. In order to recognize the object in the mobile environment, it is possible to recognize the object using the MobileNet model which is a learned model based on the deep learning. The object recognition service works with Amazon's Alexa service, which provides speech recognition and voice conversion services to provide voice services for recognized objects. In this paper, we propose a base architecture that can provide services in three aspects based on object recognition and voice service. The first one is the voice service for the recognized object, the second is the personalized voice service for the recognition object, and finally the knowledge service for the recognition object using the Alexa service.
  • 5.

    Development of Apparel Service Distribution Pattern Prediction Application Service Using Big Data

    YOONJIHAE | Kang Sun-Kyung | 2018, 13(4) | pp.451~458 | number of Cited : 0
    Abstract PDF
    There are more companies and start-ups providing recommended services to consumers by predicting and analyzing the distribution patterns under circumstances through big data of various distribution fields. This paper is to use the distribution big data of clothing to develop the distribution pattern prediction service. As the actual distribution data of clothing distributors are used for product data, import and export data, and import date data, the accuracy of distribution pattern of clothing may be enhanced. After filtering the actual distribution data of clothing distributors under items, months, and seasons by using the Hadoop-based R, the most distributed items ranked from one to ten are shown on the Android-based application. The application was developed to arrange and filter the data by category to indicate the useful data for retailers of clothing. Therefore, through the application developed, difficult data flows and distribution are easily accessible. The distribution pattern prediction application service developed in this paper shows the result of prediction of the clothing distribution patterns by circumstances (month or season) to wholesalers and retailers of clothing within the scope of data by current clothing distributors to assist the smooth distribution service.
  • 6.

    Analysis for Low Interference Effect between Multi-Antennas Using Magneto-Dielectric Material Sheets

    Yong Jin Kim | 2018, 13(4) | pp.459~464 | number of Cited : 1
    Abstract PDF
    In 2018, the world's first 5G-based pilot service was launched at the Pyeongchang Winter Olympics, and 5G technology is expected to lead the next mobile communication technology. 5G technology requires a transmission speed between 100Mbps and 1Gbps. To meet these requirements, mmWave technology, carrier aggregation technology, and MIMO (Multi Input Multi Output) antenna technology should be applied. The products in which 5G technology will be implemented are portable mobile devices such as smart phones or smart watches. In case of 5G smartphone, the antennas for 5G frequency bands should be added in addition to the existing antennas for cellular bands. However, even if 5G band antennas are additionally applied, there is little change in the external size of the mobile device. Therefore, more antennas must be mounted in the same mounting space, so it is difficult to meet the requirements of 5G technology without increasing the degree of integration. In addition, recently introduced smart phones are designed with a metal frame chassis, which makes it difficult to secure antenna mounting space and radiation performance. Therefore, various researches have been continued to overcome the radiation performance and antenna density of the 5G smartphone. In this paper, we analyze the coupling characteristics between multiple antennas using magnetic dielectric sheets through simulations on mobile devices such as smart phones, and improve the antenna density and radiation performance through the interference reduction effect shown in the analysis results.
  • 7.

    A Development of Android-based Chinese Characters Writing Recognition Algorithm

    LEE SUNG HYUN | Lim Yong Gyun | 김영철 | 2018, 13(4) | pp.465~473 | number of Cited : 0
    Abstract PDF
    Chinese character education uses the method of writing and practicing according to the letter rather than learning the meaning and the sound of the word. When making Chinese character education application, therefore, it is necessary to deal with the technical method of recognizing and writing according to the Chinese characters for increasing the effectiveness of education. In this study, we try to find an effective method for writing Chinese character based on the smart-phone device. In order to recognize Chinese characters, first, a standard sample character is recognized. Because it is a font, it is necessary to make the size and shape constant regardless of the font or the form. The next step is to recognize the user-created image and compare it with the criteria. After the comparison, it is a process of evaluating how the reference value is similar to the value created by the user. After going through three steps, we compare the reference value with the user's writing. In this study, among the existing recognition methods, DP matching and HMM methods that are easy to apply to applications and proposed algorithms that combine DP matching with HMM. The proposed algorithm has resulted in a higher recognition rate of at least 1.05% compared to the previous recognition method.
  • 8.

    Design and Implementation of IoT based Real-time Monitoring Platform for ICU

    Byoung-Chan Jeon | Jung-Hyen Cho | Da-Hyun Kang and 1other persons | 2018, 13(4) | pp.475~485 | number of Cited : 0
    Abstract PDF
    As medical and sensor technologies have recently developed, interest in the health care industry is growing. Advances in cloud computing technology enable medical data to be shared over the network. Regardless of the location of the medical team, the medical information can be shared between agencies such as hospitals and insurance companies, as well as remote care and collaboration. In addition, the storage and management of data is becoming important because health care data, which only happens when you go to a hospital and see a doctor, is growing exponentially. On the other hand, except for large hospitals, medical sensor data monitoring the patient's life-sustaining status in the intensive care unit of small hospitals may be lost without collected or stored.  In this thesis, a platform is designed to integrate medical data continuously generated from the hospital's intensive care unit into IoT environment. The IoT bridge module is developed to integrate medical sensors into the IoT environment, and medical data generated by the MQTT protocol is sent as an integrated message to the medical team's client devices and servers. The transmitted data is monitored on real-time graphs, stored in the server's storage, and designed and implemented for further analysis.
  • 9.

    Design and Implementation of Butler Matrix Based on Multilayered MTL

    Dong Hak Lee | Doo-Yeong Yang | 2018, 13(4) | pp.487~498 | number of Cited : 1
    Abstract PDF
    The proposed Butler matrix is capable of controlling the beam direction of array antenna as a feeding system of the beamforming network. The beam pattern of array antenna is controlled by the magnitude and phase of feeding signal, and interval between antenna elements. Therefore, the magnitude and phase of feeding signal supplied to array antenna is used for adjusting the beamforming direction. Designing each component such as the directional coupler, phase shifter and crossover is very important because those decides on the performance of Bulter matrix. In this paper, we propose the 4⨉4 Butler matrix combining three-branch line directional couplers with slot-coupled directional couplers to extend the limited narrow bandwidth, and to reduce the size. Therefore, the Butler matrix circuit is constructed in a multilayer technology and designed without using a crossover on the same layer. The multilayered composition is resolved by using the slot-coupled directional coupler. Finally, the Butler matrix was designed and fabricated on double layered microstrip transmission line (MTL) by the proposed configuration. And also, the simulation results and the measured results were compared for performance analysis each other. As a result, the proposed matrix has good performance that the relative bandwidth, maximum return loss, insertion loss and progressive phase shift are 24%, 15.76dB, 6.5±0.8dB and 9.5° in operating frequency band, respectively.
  • 10.

    Studies on the Antimicrobial and Anti-inflammatory Effects of the Orostachys japonicus Peptide Fractions

    Lee, Junwon | 2018, 13(4) | pp.499~507 | number of Cited : 0
    Abstract PDF
    Orostachys japonicus, a medicinal plant, is a perennial herbaceous plant that grows on rocks and tile roofs. It grows like a pine tree, so it is sometimes called a houseleek. Recently, the extract of Orostachys japonicus has been reported to have immune enhancement, anticancer effect, apoptosis inducing effect on leukemic cells, antioxidant effect, antibacterial effect, dietary anti-hyperlipidemic effect, anti-diabetic activity effect, and antimutagenic effect. The substances isolated by solvent extraction were known flavonoids, triterpenes, sterols, derivatives of quercetin and kaempferol. Protein peptide fractions were separated and analyzed for anti-inflammatory activity and antimicrobial activity. To confirm the cytotoxicity of the separated peptide 3 KDa fractions, Raw 264.7 cells were used as the macrophage cell line and the cell viability was 90% or more at a concentration of 1 μg/mL and a concentration of 10 μg/mL. As a result of measuring NO reduction rate, NO decreased about 81% at the concentration of 10 ug/mL of peptide 3 KDa fraction. The antimicrobial activity of the peptide fractions inhibited the growth of Escherichia coli, which is a gram negative bacterium, and Bacillus cereus and Bacillus subtilis, which are Gram positive bacteria, but did not exhibit strong antimicrobial activity as compared with control methanol. In the case of the products using the protein peptide, there is no research on the efficacy for the standardized manufacturing method and the origin of country. Therefore, functional studies are urgently required and this paper can be used as a basic data.