Machine Learning for Joint Channel Equalization and Signal Detection
Lin Zhang, Lie‐Liang Yang · 2019
This chapter focuses on machine learning-based channel equalization and data detection, which utilize deep learning neural networks (NNs) to learn and formulate the feature sets of time-varying wireless channels in order for them to be efficiently operated in highly dynamic wireless channels. It first presents a brief overview of ML-based equalizers, and describes three classic equalization algorithms: zero-forcing equalization, minimum mean-square error equalization, and maximum likelihood sequence estimation equalization. Next, the chapter briefly introduces a quintessential model for the feedback neural network, i.e. the multilayer perceptron (MLP), in order to provide some insight into the construction and training of NNs. Then, based on the MLP and deep learning (DL), a DL neural network solution is proposed for the implementation of channel equalizers. The chapter also investigates the performance of orthogonal frequency-division multiplexing systems involving NN-based channel equalization.