Pilot Pattern Design for Deep Learning-Based Channel Estimation in OFDM Systems

In this letter, we present a downlink pilot design scheme for Deep Learning (DL) based channel estimation (ChannelNet) in orthogonal frequency-division multiplexing (OFDM) systems. Specifically, in the proposed scheme, a feature selection method named Concrete Autoencoder (ConcreteAE) is used to find the most informative locations for pilot transmission. This autoencoder consists of a concrete layer as the encoder and a multilayer perceptron (MLP) as the decoder. During the training, the concrete layer selects the most informative pilot locations, and the decoder reconstructs an approximate estimation of the channel. Eventually, the ChannelNet is trained on the output of the ConcreteAE aiming to reconstruct the ideal channel response. The estimation error results show that this approach outperforms the previously presented ChannelNet with a uniformly distributed pilot pattern, and its performance is comparable to the minimum mean square error (MMSE).

I. Introduction

OFDM modulation has been adopted widely in wireless networks such as LTE and 5G NR. To guarantee the reliable signal detection, it is necessary to efficiently estimate the channel state information (CSI) at the receiver. Pilot-based channel estimation is one of the most common methods for obtaining the CSI. Pilots are some symbols placed in specific locations of the time-frequency grid which their positions and values are known for the receiver. The receiver then estimates the channel response in all of the time-frequency grid based on the channel it observes on the neighboring pilot tones. There are some conventional methods conducted for channel estimation in OFDM systems like least square (LS) and MMSE. The LS method is an interpolation-based approach while the MMSE has a better performance since it uses the complete channel statistics; however, it suffers from higher computational complexity and needs prior knowledge of the channel statistics.