Extreme learning machine and correntropy criterion-based hybrid precoder for 5G wireless communication systems

Extreme learning machine and correntropy criterion-based hybrid precoder for 5G wireless communication systems

연구 분야: Networking

학회: Signal, Image and Video Processing

Application of massive multiple input multiple output (mMIMO) in millimeter wave (mmWave) band is a promising solution for 5G communication due to low latency and directional beamforming. Hybrid precoding is an integral part of 5G systems to fully exploit spatial information in presence of large path loss with reduction of RF chains. However, hybrid precoder design is a challenging research concern due to the involvement of large number of antennas in transmitter and receiver. Additionally, higher spectral efficiency in the presence of impulsive noise in FR2 or mmWave band also requires attention. ELM (extreme learning machine) has an efficient feed forward neural architecture with only one hidden layer which is suitable to design efficient precoding models. Therefore, this research focuses on to design hybrid precoder in millimeter wave band using ELM and variable center correntropy criterion to obtain higher spectral efficiency. Exhaustive simulation results indicate that the proposed precoder has significantly better performance than state-of-art methods in multiple test conditions. The precoder performance is tested by varying base station antennas, mobile station antennas and number of users. The bit error rate (BER) performance is also analyzed. and comparison results are presented to justify the claim.