Millimeter Wave (mmWave) and massive MIMO (Multiple Input Multiple Output) are promising solutions for 5G communications. Hybrid precoding architecture (analog and digital) is generally employed to resolve high hardware complexity and energy consumption issues. The current hybrid precoding architectures are computationally complex. This proposes a novel deep neural network based precoding architecture named ‘ComcepNet’. The network combines the features of Complex Convolution blocks and Inception Network. The network is observed to deliver superior performance in terms of accuracy and achievable datarate compared to the present Autoprecoder network.
Paper
Full text
Deep Learning based Hybrid Precoding for mmWave Massive MIMO system using ComcepNet
Semantic Scholar · Engineering · 2020
Abstract
Millimeter Wave (mmWave) and massive MIMO (Multiple Input Multiple Output) are promising solutions for 5G communications. Hybrid precoding architecture (analog and digital) is generally employed to resolve high hardware complexity and energy consumption issues. The current hybrid precoding architectures are computationally complex. This proposes a novel deep neural network based precoding architecture named ‘ComcepNet’. The network combines the features of Complex Convolution blocks and Inception Network. The network is observed to deliver superior performance in terms of accuracy and achievable datarate compared to the present Autoprecoder network.