Convolutional fuzzy neural network based symbol detection in MIMO NOMA systems

One of the most important tasks to be considered in wireless communication systems, especially in multi-carrier systems such as Multi-Input Multi Output Non-Orthogonal (MIMO-NOMA), is to correctly estimate the channel state information for coherent detection at the receiver. A hybrid deep learning model, called convolutional fuzzy deep neural networks, is proposed in this study for accurately estimating channel state information and detecting symbols in MIMO-NOMA systems. The performance of this proposal has been compared to traditional algorithms like Least Square Error- Successive Interference Cancelation (LS-SIC) and linear minimum mean square (LMMSE-SIC), as well as to other deep learning methods such as convolutional neural networks. With this proposed scheme, significantly less bit error rate is obtained in both Rician and Rayleigh channel environment compared to other algorithms. In addition to the high performance of this scheme, the fact that it does not need channel statistics is another important advantage.