MaIC-Net: Learnale weight scabling and complexity scalable deep neural network for massive mimo detection

Abstract

Massive multiple-input multiple-output (MIMO) is a crucial method for next-generation communication systems that must satisfy user performance as well as quality of service (QoS) standards. A vast MIMO uplink receiver’s symbol analyzers became increasingly complicated as the number of antennas and radio frequency (RF) chains increased. Thus, research into the ideal massive MIMO detection method with high efficiency as well as minimal complexity has garnered a lot of attention through the last decade. Therefore, this paper presents a Multiuser Accelerated Interference Cancellation Network (MaIC-Net) to perform massive MIMO detection in the uplink communication system. In addition, a monotonic non-increasing profile function is designed to scale each layer presented in the MaIC-Net, allowing the proposed MaIC-Net model to dynamically learn the attenuation function. In addition, the sparsity-inducing regularization requirement is combined with accessible weight scaling methods to decrease model size and computational complexity. Similarly, the SoftS activation function is intended to enhance detection results by higher-order modulation. The Matlab tool is used to execute the proposed model and evaluate the performance of the proposed mode using various performance parameters. The result analysis reveals that the proposed method outperforms existing methods by achieving 0.000319 less symbol error rate (SER) against 8dB Signal-to-Noise-Ratio (SNR) for the 64-Quadrature Amplitude Modulation (QAM) modulation technique.