Deep-Neural-Network Based Fall-Back Mechanism in Interference-Aware Receiver Design

In this paper, we consider designing a fall-back mechanism in an interference-aware receiver. Typically, there are two types of detectors dealing with interference, known as enhanced interference rejection combining (eIRC) and symbol-level interference cancellation (SLIC). Although a SLIC detector performs better, it yields a higher complexity than an eIRC detector. Further, it requires knowledge of interference modulation-format (MF). Due to potential detection errors, SLIC can run with a wrong interference MF and render unsatisfying results. Therefore, designing a mechanism that runs SLIC when the interference MF is reliable and otherwise switches to eIRC is of particular interest, which we call a "fall-back mechanism". Finding an optimal mechanism is difficult and we use deep-neural-network (DNN) for design, which is more effective than a traditional Bayes-risk mini-mization based approach.

1. INTRODUCTION

In modern communication systems, multiple users and spatial data-streams are multiplexed on the same time-frequency resources to increase spectral-efficiency. This brings challenges in recovering useful data from interfered signals. In an interference-aware receiver, two detectors are widely considered, namely, enhanced interference rejection combining (eIRC) and symbol-level interference cancellation (SLIC) [1]. SLIC reconstructs and removes interference from the corrupted signal before detecting useful signal, while eIRC only suppresses interference instead of canceling it [2], [3]. SLIC is more advanced, but at a cost of higher complexity, less robustness, and it requires more prior information.