1
Machine Learning Aided Channel Equalization in Filter Bank Multi-Carrier Communications for 5G
Ubaid M. Al-Saggaf1,2, Muhammad Moinuddin1,2*, Syed Saad Azhar Ali3, Syed Sajjad Hussain Rizvi4 and Muhammad Faisal5
1Center of Excellence in Intelligent Engineering Systems (CEIES), King Abdulaziz University, Jeddah, Saudi Arabia
2Electrical and Computer Engineering Department, King Abdulaziz University, Jeddah, Saudi Arabia
3Center for Intelligent Signal and Imaging Research (CISIR), Electrical and Electronics Engineering Department, Universiti Teknologi PETRONAS, Bandar Seri Iskandar, Malaysia
4Computer Science Department, SZABIST, Karachi, Pakistan
5Computer & Information Technology Dept., Dammam Community College, King Fahd University of Petroleum & Minerals, Dhahran, Saudi Arabia
Abstract
Multi-carrier communications (MC) have gained a lot of interest as they have shown better spectral efficiency and provide flexible operation. Thus, the MC are strong candidates for the fifth generation of mobile communications. The Cyclic-prefix orthogonal frequency division multiplexing (CP-OFDM) is the most famous technique in the MC as it is easy to implement. However, the OFDM has poor spectral efficiency due to limited filtering options available. Thus, to enhance spectral efficiency, an alternative to OFDM called Filter bank multicarrier (FBMC) communication was introduced, which has more freedom of filtering options. On the other hand, the FBMC preserves only real orthogonality for the waveforms, resulting in imaginary interference. Hence, the equalization in FBMC has to deal with this additional interference which becomes challenging in multiuser communication. In this chapter, the aim is to deal with this challenge.
Keywords: Multiuser communications, multicarrier communications, OFDM, FBMC, 5G, equalization, machine learning, MMSE
1.1 Introduction
Improved bandwidth, efficient power utilization, and better handling of selective fading are prominent advantages of the MC [1]. The CP-OFDM is a simpler MC solution. However, it has poor spectral efficiency due to limited filtering options available.
Another candidate considered for multiple radio access in 5G is the Filter Bank Multi-Carrier (FBMC) [2]. The FBMC is a modified version of OFDM, where the rectangular transmit and receive waveforms are replaced by any other more frequency localized pulse shapes, which offer a more confined spectrum, relevant for spectrum-sharing scenarios.
In the FBMC with Offset Quadrature Amplitude Modulation (OQAM) system, Channel equalization is challenging compared to the standard CP-OFDM. This is because the real and the imaginary symbols in the FBMC-OQAM system are transmitted with a time offset, which results in loss of orthogonality in the imaginary part [2]. Thus, the received signal suffers from the additional imaginary part, which is termed as Imaginary interference. This interference degrades the performance of the channel estimator. As a result, channel equalization also degrades as it requires the knowledge of channel estimates. In conventional equalization methods for the FBMC-OQAM, the channel estimates are used to cancel the phase of the received signal and extract its real part. However, due to errors in channel estimates, the phase cancellation is not completely achieved. Thus, this conventional approach does not work in this scenario. Alternatively, people have proposed direct equalization methods, which still need improvements. Therefore, there is a need to design an efficient equalization method to combat the imaginary interference and inter-symbol interference (ISI), and inter-carrier interference (ICI) in the FBMC-OQAM system.
1.2 Related Literature Review
The FBMC modulation is a recent type of MC technique develop...
