Channel Estimation of Massive MIMO-OFDM System Using Elman Recurrent Neural Network

Abstract In a cell-free massive MIMO system, multiple users arrive at multiple access points at separate times, while in an OFDM system, different delays can be equivalent to symbol timing offsets (STOs). Since symbol timing offsets are not all the same, in the downlink transmission process, it is necessary to consider its impact on transmission techniques, such as channel estimation and downlink precoding. In this paper, aiming at the performance loss caused by STO in cell-free massive MIMO-OFDM system, we propose a multi-RB precoding optimization algorithm that maximizes the downlink sum rate. We derive the sum rate maximization problem into an iterative second-order cone programming (SOCP) form to achieve convex approximation. Then, considering the impact of STO on the accuracy of cell-free massive MIMO-OFDM channel estimation, we propose a downlink channel estimation method, which jointly uses channel state information reference signal (CSI-RS) and demodulation reference signal (DMRS). Simulation results show that the proposed multi-RB optimal precoding can effectively improve the downlink sum rate, and the proposed downlink channel estimation can obtain accurate multi-RB frequency domain channel parameters.

Download Full-text

SoftMax Recurrent Neural Network based Faded Channel Estimation using FIM in MIMO-OFDM

SMART MOVES JOURNAL IJOSCIENCE ◽

10.24113/ijoscience.v6i2.265 ◽

2020 ◽

Vol 6 (2) ◽

Author(s):

Gulzar Ansari ◽

Dr. Abhishek Bhatt

Keyword(s):

Neural Network ◽

Channel Estimation ◽

Recurrent Neural Network ◽

Least Squares Method ◽

Time Series Prediction ◽

Research Work ◽

Estimation Technique ◽

Ongoing Research ◽

Estimation Techniques ◽

Mimo Ofdm

The combination of MIMO technology with OFDM system, there is enhancement of wireless digital communication which is quite beneficial for future communicating system. MIMO-OFDM improves the efficiency and quality of the wireless scenario. With efficient channel estimation technique especially non-blind under MIMO-OFDM scenario present an enhanced performance with low complexity. In the pilot type, the least squares method (LS method) is less complex and requires an implicit knowledge of the channels. However, it suffers from inter-carrier interference (ICI). For this reason, the optimal design of Channel Estimator is an area of ongoing research. In this work, the performance of the DWT-OFDM scheme in combination with a multi-input system and multiple outputs for unknown pilot symbols is evaluated using a neural network. Take advantage of time series prediction using a recurrent neural network (RNN) with a SoftMax layer with frequency index modulation to perform channel prediction. In this research work a comparative analysis is performed among proposed softmax recurrent neural network based channel estimation with existing channel estimation technique for variable signal to noise ratio (Eb/No) for frequency indexed modulation techniques. The existing channel estimation techniques for MIMO-OFDM communicating environment are based on known pilots over the noisy fading wireless environment. From simulation result, it is observed and concluded that the existing channel estimation techniques gives higher BER as compared to proposed softmax recurrent neural network channel estimation technique.

Download Full-text