Machine Learning Detectors for MU-MIMO Systems with One-bit ADCs

Abstract

We consider an uplink multiuser multiple-input multiple-output (MU-MIMO) system with one-bit analog-to-digital converters (ADCs). In this system, the construction of a low-complexity detector is quite challenging due to the non-linearity of an end-to-end channel transfer function. Recently, a supervised-learning (SL) detector was proposed by modeling the complex non-linear function as a tractable Bernoulli-mixture model. It achieves an optimal maximum-likelihood (ML) performance, provided the channel state information (CSI) is perfectly known at a receiver. However, when a system-size is large, SL detector is not practical because of requiring a large amount of labeled data (i.e., pilot signals) to estimate the model parameters. We address this problem by proposing a semi-supervised learning (SSL) detector in which both pilot signals (i.e., labeled data) and some part of data signals (i.e., unlabeled data) are used to estimate them via expectation-maximization (EM) algorithm. We further extend the proposed detector for time-varying channels, by leveraging the idea of online learning, which is called online-learning (OL) detector. Simulation results demonstrate that the proposed SSL detector can achieve the almost same performance of the corresponding SL detector with significantly lower pilot overhead. In addition, it is shown that the proposed OL detector is more robust to channel variations compared with the existing detectors.