In this paper, we present a deep learning-based technique for channel estimation. By treating the time-frequency grid of the channel response as a low-resolution 2D-image, we propose a 5G-New Radio Convolutional Neural Network, called NR-ChannelNet, which can be properly trained to predict the channel coefficients. Our study employs a 3GPP-compliant 5G-New Radio simulator that can reproduce a realistic scenario by including multiple transmitting/receiving antenna schemes and clustered delay line channel model. Simulation results show that our deep learning approach can achieve competitive performance with respect to traditional techniques such as 2D-MMSE: indeed, under certain conditions, our new NR-ChannelNet approach achieves remarkable gains in terms of throughput.
A deep learning-based approach to 5G-new radio channel estimation / Zimaglia E.; Riviello D. G.; Garello R.; Fantini R.. - ELETTRONICO. - (2021), pp. 78-83. (Intervento presentato al convegno 2021 Joint European Conference on Networks and Communications & 6G Summit (EuCNC/6G Summit) tenutosi a Porto, Portogallo nel 8-11 June 2021) [10.1109/EuCNC/6GSummit51104.2021.9482426].
A deep learning-based approach to 5G-new radio channel estimation
Riviello D. G.
;
2021
Abstract
In this paper, we present a deep learning-based technique for channel estimation. By treating the time-frequency grid of the channel response as a low-resolution 2D-image, we propose a 5G-New Radio Convolutional Neural Network, called NR-ChannelNet, which can be properly trained to predict the channel coefficients. Our study employs a 3GPP-compliant 5G-New Radio simulator that can reproduce a realistic scenario by including multiple transmitting/receiving antenna schemes and clustered delay line channel model. Simulation results show that our deep learning approach can achieve competitive performance with respect to traditional techniques such as 2D-MMSE: indeed, under certain conditions, our new NR-ChannelNet approach achieves remarkable gains in terms of throughput.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.