A Survey on Learning-Based Channel Estimation Methods Used for 5G Massive Mimo System

The continuous evolution in wireless communication, including fifth-generation (5G), tends to allow lower latency, higher data rates, and more energy efficiency. Fifth generation(5G) technology requires a highly precise approximation of the ‘channel state information' (CSI) to utilize the advantages of ‘massive multiple input multiple output' (Ma-MIMO) systems. Ma-MIMO has raised excessive interest in the industry of 5G communication, as it can efficiently raise the capacity of the communication system and lowering the transmitting power. In Ma-MIMO, 5G ‘orthogonal frequency division multiplexing' (OFDM) systems use pilot strings to identify channel behavior by following one of the traditional methods such as ‘least squares' (LS) or ‘minimum mean square error' (MMSE) estimation. However, in case of insufficient channel parameters (i.e., low signal to noise ratio), LS gives a poor estimation and MMSE requires statistical cognition of the channel parameters and noise in advance. This paper presents various approaches of deep learning-based framework used in deep neural network (DNN) architecture designed for 5G Ma-MIMO channel estimation.