Deep Learning-Based Model for Simultaneous Channel Estimation and Signal Decoding for Wireless Communication

In recent decades the world has witnessed significant demand for the Internet facility due to the evolution of cellular communication from 1G to 5G technology. The intensive research in wireless technology has shown that the inception of Non-Orthogonal Multiple Access (NOMA) provides unprecedented high spectral and energy efficiency, thereby supporting massive connectivity. To pave the way for attention towards intelligent systems in wireless communication systems, this proposed work uses a Deep Learning (DL) approach. Here, DL is used implicitly to perform channel estimation and signal decoding simultaneously for Orthogonal Frequency Division Multiplexing (OFDM) based NOMA systems. This work explores the potential of DL to combat decoding complexity faced by the conventional Successive Interference Cancellation (SIC) technique. It also ensures the superiority of DL over usual methods such as Least Square (LS) and Minimum Mean Square Error (MMSE) channel estimation techniques considering two users per cluster in a cell region. It is observed that the use of DL for user 1 has resulted in 18.75% enhancement over MMSE and 25% over LS. Neural networks with different DL-based sequential layers such as Gated Recurrent Unit (GRU), Long short-term memory (LSTM), and Bi-directional Long short-term memory (Bi-LSTM), along with different optimizers like Adaptive Moment Estimation (ADAM), Stochastic Gradient Descent with Momentum (SGDM), and Root Mean Square Propagation (RMSprop) are tested to verify the proposed model. Further, the simulation results of conventional channel estimation techniques combined with SIC decoding technique are compared with the proposed DL model using symbol error rate (SER) performance.