Bi-LSTM Based Deep Learning Algorithm for NOMA-MIMO Signal Detection System

The bi-directional long short-term memory (Bi-LSTM) is regarded as one of the most promising deep learning algorithms for signal detection in Multiple inputs and Multiple outputs (MIMO) system. The projected algorithm efficiently estimates the channel for several multipath scenarios and enhanced the throughput of system. The simulation results reveal that the projected Bi-LSTM offer a significant Bit error rate (BER) improvement as compared with LSTM, zero forming equalizer (ZFE), minimum mean square error (MMSE) and maximum likelihood (ML) methods for 8 x 8,16 x 16 and 64 x 64 MIMO system with Rayleigh channel. In the proposed work, the deep learning algorithm based Bi-LSTM method is explored for signal detection in 5G for NOMA MIMO systems. The projected Bi-LSTM is compared with the conventional methods by estimation the BER and complexity of the detection algorithms. The simulation results demonstrate that the Bi-LSTMM gave a superior efficiency as compared with LSTM, MMSE, ML and ZFE schemes.