Blind identification of MIMO Space-Time Block Codes Based on Convolutional Neural Network

Abstract

Aiming at the blind identification of space-time block codes (STBC) in multiple input multiple output (MIMO) system, this paper proposes a new convolutional neural network (CNN-N) to realize the blind identification. Compared to traditional algorithms using statistical features of received signal, CNN-N can reduce the computation with a higher correct identification rate. Consist of multiple layers with special functions, CNN-N has good generalization ability in complex MIMO channels. The network in this paper can recognize 6 STBC codes include spatial multiplexing signal (SM) and some OSTBC codes. The simulation result shows that this new convolutional neural network can finish STBC identification with a high correct rate even in low SNR by consuming moderate amounts of time.