A Massive MIMO detection method based on deep learning

• 1、School of Information and Communication Engineering, Beijing University of Posts and Telecommunications, Beijing 100876

Abstract：Massive multiple input and multiple output(Massive MIMO) detection technology is a research hotspot in spectrum sensing. It is a key technology in modern wireless communication systems. At present, detection methods mostly use rule-based heuristic algorithms, which cannot model the channel information. This paper applies deep learning to signal detection in massive MIMO systems. There are many detection algorithms now which could not be applied in actual communication systems due to the high complexity. On the other hand, their performance cannot achieve optimal performance, so they cannot meet the requirements of modern wireless communication systems such as 5G for spectral efficiency. Therefore, a massive MIMO detection algorithm based on deep learning is proposed, that is Optimized Sparsely Connected Network(O-ScNet). Compared with prior deep learning algorithms, the proposed detection algorithm O-ScNet optimizes input features and loss functions. The simulations results show that the O-ScNet algorithm can converge faster and achieve better experimental results.

Keywords： Massive MIMO deep learning residential network loss function