Deep Learning Based Massive-MIMO Decoder

Massive Multiple Input Multiple Output (M-MIMO) is one of the key technology for 5
^th
generation (5G) communication systems. Designing of M-MIMO decoder is very complex and challenging task. Recently, deep neural network (DNN) emerges as an alternative way to perform many complex tasks. In this work we have presented a deep learning (DL) based architecture DLNet) for receive signal decoding in Massive-MIMO system. In this we have considered an uplink time-varying Gaussian random MIMO channel perfectly known to the receiver. Based on the knowledge of this MIMO channel and the received signal the proposed DLNet decoder is decoding the messages of all the users. The presented decoder is 50 layer deep neural network. Its architecture is based on the projected gradient descent algorithm. Our simulation result shows, the proposed DLNet decoder for M-MIMO system performs better than other MIMO decoding techniques, by 3 dB bit error rate (BER), atleast 124 times faster and 9 times less complex.