본문 바로가기 주메뉴 바로가기
22대 대한민국 국회 국회정보길라잡이 국회의원검색
회원가입 로그인
HOME

정보검색

소장정보 검색

국회도서관 홈으로 정보검색 소장정보 검색
결과 내 검색 동의어 포함
결과 내 검색 동의어 포함

-

목차보기

Title Page

Abstract

Contents

I. Introduction 8

1.1. Main Contributions 8

II. Learning-based Wireless Communications with Energy Harvesting 10

2.1. Downlink Beamforming in Small Cells with Scalar Information Exchange Introduction 10

2.2. Min-SINR Maximization with DL SWIPT and UL WPCN in Multi-Antenna Interference Networks 16

2.3. RNN-Based Node Selection for Sensor Networks with Energy Harvesting Introduction 26

III. Learning-based Robot Vision Systems 33

3.1. Privacy-Preserving Robot Vision with Anonymized Faces by Extreme Low Resolution 33

References 47

Figure 1. System model 11

Figure 2. Achievable sum-rate/cell vs. SNR for NC＝3 and NT＝2[이미지참조] 15

Figure 3. Proposed joint time switching SWIPT and WPCN protocol 17

Figure 4. DL beamforming with energy harvesting and UL power allocation in multi-cell MISO networks 18

Figure 5. Proposed max-min-SINR DL beamforming and UL PA design 23

Figure 6. Minimum UL user rate vs. minimum DL user rate 24

Figure 7. θmin, θmax vs. number of iterations[이미지참조] 25

Figure 8. Superframe structure 26

Figure 9. Basic structure of RNN 27

Figure 10. System model 28

Figure 11. UL packet format 29

Figure 12. RNN structure of the proposed scheme 29

Figure 13. Method to make labeled ground truth 30

Figure 14. Training loss vs. Learning iterations 31

Figure 15. Number of penalties vs. time 31

Figure 16. Composition of the developed patrol robot system with privacy preserving face detection. 35

Figure 17. Dynamic resolution face detection architecture. 37

Figure 18. Our training data generation process for the example of the image with two faces. 39

Figure 19. Results comparing our proposed method with the results of the approach presented... 43

Figure 20. Face detection robot used in our experiments 44

Figure 21. Comparison of the feature extraction results at various resolutions. 45

초록보기

From self-driving cars to smartphones essential to our lives, many types of the electronic devices and computers handle intelligently our work. Thanks to the 'things' that have become smarter, our lives have become more pleasant and faster, and literally easier. One of the big reasons we can live in such an environment is 'machine learning'. It is a technology that allows a machine to acquire new knowledge by learning through a huge amount of data, just like a person learns. Machine learning is one of the most important topics in many industries and researches these days. It is no exaggeration to say that machine learning is used in almost every field. Its application to (1) wireless communications and (2) computer vision based robotics are also essential.

Learning based communication system has the following possibilities: (1) Unlike communication theory, real communication systems are non-linear. For this reason, deep learning-based communication systems may be more suitable for specific hardware configurations and channel optimization. (2) One of the great features of a communication system is that various signal processing functions (e.g., Coding, modulation, detection) are separated into several blocks. Rather than optimizing the performance of individual blocks, a machine learning-based end-to-end communication system can perform better. Because of these possibilities, machine learning is being applied to a wide range of communication systems such as heterogeneous access technology, cognitive radio, and resource allocation. In this dissertation, we propose a mathematical approach to the optimization problem of interference mitigation in a multi-cell network with and without energy harvesting. Also, we propose a recurrent neural network (RNN) based node selection algorithm for sensor networks with energy harvesting. Comparing the problem solving method of the former and the latter, the difference between the existing communication system and the learning based communication system can be clearly revealed.

Computer vision based robotics is a study that extracts meaningful information from an image or video and applies the information to a robot. In particular, as a result of applying machine learning to this field, various robots, such as autonomous vehicles, unmanned courier robots, and smart home robots, are being developed. The more studies on robots equipped with cameras, the more convenient our lives, but on the contrary, they can invade our privacy. That is, it is a double-edged sword. In this dissertation, we propose a method to protect our privacy while utilizing other visual information well (i.e., Simultaneous localization and mapping (SLAM)) by detecting faces in extreme low resolution images.

추천서가 (다양한 추천 자료를 만나보세요)

권호기사보기

권호기사 목록 테이블로 기사명, 저자명, 페이지, 원문, 기사목차 순으로 되어있습니다.
기사명 저자명 페이지 원문 기사목차