Multilayer Perceptron Scheme for Beamforming and Channel Estimation of Massive MIMO
Sneha V.V1, Ismayil Siyad C2, S.Tamilselvan3

1Sneha V. V, PG Scholar, Department of ECE, MEA Engineering College, Perinthalmanna (Kerala), India.
2Ismayil Siyad C, Assistant Professor, Department of ECE, MEA Engineering College, Perinthalmanna (Kerala), India.
3S. Tamilselvan, Associate Professor, Pondicherry Engineering College, (Puducherry), India.
Manuscript received on 19 August 2019 | Revised Manuscript received on 29 August 2019 | Manuscript Published on 16 September 2019 | PP: 467-471 | Volume-8 Issue-2S6 July 2019 | Retrieval Number: B10890782S619/2019©BEIESP | DOI: 10.35940/ijrte.B1089.0782S619
Open Access | Editorial and Publishing Policies | Cite | Mendeley | Indexing and Abstracting
© The Authors. Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open access article under the CC-BY-NC-ND license (

Abstract: Massive MIMO being one of the core enabler of 5G network, estimation of channel characteristics in such network plays a vital role. Also the channel is most affected by the phase information rather than the amplitude information. Thus phase estimation plays a very important role in order to ensure error free transmission especially in the case of rapidly changing channels like in massive MIMO. When an accurate transmit beamforming technique is employed to this scheme, it results in maximum user data separation thereby improving the directivity of the signal. The proposed scheme works on two multilayer perceptron (MLP) mechanism model, one for beamforming and other for channel estimation. First MLP model takes the transmitted signal that is passing through the channel to generate the beamforming vectors and these beamforming vectors after prediction are then multiplied with the channel coefficients to enable beamforming. The transmit beamformed signal after passing through the channel and applying appropriate noise, will be received in different directions and these training samples are then given to the second MLP model to predict DOA of the received signal which in turn estimate the channel of massive MIMO. In our proposed scheme, as both beamforming and channel estimation are handled by deep neural network, along with achieving very much better accuracy, mean square error (MSE) and bit error rate (BER) performance, it reduces the number of epochs required for training which results in an efficient learning scheme that is capable of predicting real time rapidly varying channel of massive MIMO.
Keywords: MIMO, 5G Communication, Deep learning, Channel estimation, DOA Estimation, Beamforming.
Scope of the Article: Electrical and Electronic Engineering