Abstract
Massive multiple-input multiple-output (mMIMO) systems with 1-bit analog-to-digital converters reduce power consumption. Nonlinearity of 1-bit quantization, however, complicates the channel estimation (CE) and data detection (DD). We propose a stochastic variational inference (SVI) based receiver design for uplink 1-bit mMIMO systems which approximates the posterior distribution over data symbols and channel coefficients. The proposed SVI algorithm maximizes the evidence lower bound associated with 1-bit likelihood, and employs Monte Carlo–based gradient estimates to obtain low-complexity parameter updates. We numerically show that the SVI algorithm outperforms existing algorithms, and that too with a lower runtime.
Funding Information
  • Centre for Development of Telematics (C-DoT), Government of India (SP/2025/06)

This publication has 12 references indexed in Scilit: