Multi-antenna system channel estimation method based on deep learning

Abstract

The invention discloses a multi-antenna system channel estimation method based on deep learning, which is suitable for estimating an uplink multipath channel and is realized based on a conditional generative adversarial network, wherein the conditional generative adversarial network comprises a generator and a discriminator based on a deep learning network. The method comprises an offline training part and an online testing part: the offline training part comprises the steps: generating a training sample according to a real channel measurement value, and then acquiring an estimated channel corresponding to the training sample by utilizing a generator; enabling a discriminator to obtain the discrimination output, and calculating a loss function to update network parameters of the discriminator and the generator; after loop iteration is completed, storing the trained generator neural network at the base station; and an online test stage comprises the steps: inputting the quantized pilot signal and the original pilot signal into the trained generator to obtain estimated channels from a user to all antennas. Compared with the prior art, the normalized mean square error of estimation can be effectively reduced.

Description

technical field [0001] The invention belongs to the field of channel estimation in wireless communication, and in particular relates to a channel estimation method for a multi-antenna system based on deep learning. Background technique [0002] Massive multi-antenna (Multiple-Input-Multiple-Output, MIMO) technology is a key technology in 5G and future communications. In a massive MIMO system, the number of antennas at the base station is large, and each antenna is equipped with a high-resolution analog-to-digital converter (Analog-to-Digital Converter, ADC), which will greatly increase the energy consumption and hardware cost at the base station. However, if a low-precision ADC is used for each antenna, system performance will be degraded due to quantization distortion caused by the low-resolution ADC. A hybrid ADC structure in which only a small number of antennas are equipped with high-resolution ADCs and the rest are equipped with low-resolution ADCs can achieve a good c...

AI Technical Summary

Problems solved by technology

However, the existing deep learning channel estimation methods generally do not study the loss function, or use the more common L 1 or L 2 loss function

This type of loss function is not designed for the channel estimation problem in massive MIMO, it will limit the performance to a large extent and lead to poor channel estimation results, there is still a lot of room for improvement

Images