Attention Model for Massive MIMO CSI Compression Feedback and Recovery

In massive multiple-input multiple-output (MIMO) scenario, in order to adapt to channel features and ensure the high-reliability and high-rate of communication, the channel state information (CSI) should be sent back to the base station (BS). As the number of antennas increases, the amount of CSI feedback increases dramatically. Ensuring communication performance and reducing CSI feedback is a challenging research direction. Some deep learning (DL) based models have been proposed to solve this problem. This paper aims to improve recovery performance and decrease time complexity. First, in encoder network, long short-term memory (LSTM) network have been introduced; second, in decoder network, attention mechanism has been added; third, early stopping have been used in training step. The simulation results show that the structure proposed in this paper is not only better than the performance of the traditional compressed sensing algorithm, but also better than the existing DL-based feedback network and achieve state-of-the-art.