Please use this identifier to cite or link to this item:
Full metadata record
DC FieldValueLanguage
dc.contributor.authorZheng, Zheyuanen_US
dc.identifier.citationZheng, Z. (2021). Deep learning application for massive MU-MIMO communication system. Master's thesis, Nanyang Technological University, Singapore.
dc.description.abstractShannon's theorem stipulates the communication rate threshold of a single link in a wireless communication system. In recent decades, scientists and developers in the communication field have spared no effort to break and conquer the communication rate limit defined by Shannon's theorem, which is the intrinsic communication capacity obtained according to the SNR and communication bandwidth of the signal in the system. With the development of traditional communication engineering, some solutions had been proposed decades ago, such as multiplexing (CDMA, TDMA, and FDMA) and advanced signal processing techniques (adaptive filter). Unfortunately, these methods are keen to improve the capacity by setting the parameters on the right side of the Shannon equation instead of breaking the threshold. However, since the beginning of the 3GPP era, whether it is the subsequent evolutionary LTE system or the current 5G in the R15-R16 stage, operators have achieved ultra-high communication rates with disproportionate improvement in parameters mentioned above. This is not because the Shannon threshold was broken, but the emergence and maturity of MIMO antenna technology which turned the single link into multiple link pairs. However, this innovative and revolutionary progress based on distributed communication methods will inevitably bring about an enormous number of replacements and technical difficulties in the block-based traditional communication systems. For example, Source Coding becomes more complicated. The Turbo code widely used in 4G LTE will be more difficult to understand than CDMA in 3GPP. Therefore, this project aims to use a Neural Network which is ‘psychic’ and good at ‘reshaping’ itself when it has a large amount of data, especially for the receiver side. In this case, all the problems such as attenuation, distortion, superposition, delay, and ISI in the communication process can be properly solved by training in the particular scenario. The advanced wireless communication systems beyond 4G are dedicated to providing high-speed capacity and quality (reliability) of the communication system. Therefore, the research firstly focuses on the efficiency improvement technologies such as massive MIMO and OFDM, while next move to the signal detector which is replaced by a neural network compared to conventional one, together with the theoretical structure analysis and comparison of simulation results to ensure that a network structure with better performance is selected and applied. In this dissertation, we will show the simulation results with the comparison to the conventional BER curve, where some popular methods like Maximum Likelihood or MMSE Algorithm are applied for channel estimation and signal detection in orthogonal frequency-division multiplexing (OFDM) systems. Instead of constructing the system and recover the signal by combining several blocks with different functions, the targeted signal detector is regarded as a ‘Black Box’, which leads to a huge disparity compared to analysis on MIMO and OFDM, where the mathematical and principal analysis can be seen frequently. The spatial multiplexing can support end-to-end high data rate transmission while providing spatial diversity gain in a single-user MIMO system. Consider a fact in the practical use of wireless communication: each communication link has two terminals. In modern communication systems wireless networks, these two terminals are often the base station and the user. When they use different operating modes, the communication link switches between uplink and downlink. In this project, we assume a multi-user (MU) model with huge arrays of antennas that can be equated to complex matrices. In the MU model, user terminals are often treated as a grid structure, while each unit in the cluster shares the same content and the communication link becomes a wireless resource for multiple users.en_US
dc.publisherNanyang Technological Universityen_US
dc.subjectEngineering::Electrical and electronic engineering::Wireless communication systemsen_US
dc.titleDeep learning application for massive MU-MIMO communication systemen_US
dc.typeThesis-Master by Courseworken_US
dc.contributor.supervisorTeh Kah Chanen_US
dc.contributor.schoolSchool of Electrical and Electronic Engineeringen_US
dc.description.degreeMaster of Science (Communications Engineering)en_US
item.fulltextWith Fulltext-
Appears in Collections:EEE Theses
Files in This Item:
File Description SizeFormat 
Deep-learning Application for Massive MU MIMO - OFDM System_G2001631J_Zheng Zheyuan.pdf
  Restricted Access
3.82 MBAdobe PDFView/Open

Page view(s)

Updated on May 17, 2022


Updated on May 17, 2022

Google ScholarTM


Items in DR-NTU are protected by copyright, with all rights reserved, unless otherwise indicated.