Generic Deep Learning-Based Linear Detectors for MIMO Systems over Correlated Noise Environments

Abstract

To support the and development and application of the fifth-generation (5G) communication and internet of things (IoT) networks, high data-rate wireless transmission is required. To meet the demand of high data-rate, multiple antennas are equipped at the transmitter and receiver, forming multiple-input multiple-output (MIMO) systems. A big challenge of MIMO is the detector design in correlated noise environments, which should achieve a fine performance with moderate computational complexity. To this end, we employ an iterative framework of a deep convolutional neural network (DCNN) and a linear detector for MIMO systems over correlated noise environments. In this framework, the linear detector can be zero-forcing (ZF), minimum mean square error (MMSE), ZF with successive interference cancellation (ZF-SIC), or MMSE-SIC, which produces an initial estimate of transmitted signals. The DCNN is used to capture the local correlation among noise, and it can produce a more accurate estimate of transmitted signals. Simulation results are finally provided to show that the proposed detector can outperform the conventional linear detectors substantially through capturing the local correlation characteristics among noise.