Deep Learning Based Detection and Channel Tracking for MIMO Systems

In this paper we consider Multiple-Input-Multiple-Output (MIMO) detection and time varying channel tracking using deep neural networks. BiLSTM-DetNet(BD-Net) is considered as a nonlinear projection operator which can be optimized as the maximum likelihood estimate using the projection gradient descent method. The trained BD-Net is also used for channel tracking based on gradient descent method. The numerical results show that the proposed approaches achieve state-of-the-art performance while maintaining fast convergence and low computational complexity requirements.