On the Blind Classification of Single-and Multi-Carrier Signals

초록

In non-cooperative contexts, such as smartmobility-enabled cognitive radios and spectrum surveillance via unmanned vehicles, a receiver must blindly identify the modulation type without any prior knowledge. Especially, distinguishing between single-carrier (SC) and multi-carrier (MC) signals is an essential prerequisite for subsequent signal demodulation and data recovery. Accordingly, a wide range of studies have been conducted to blindly classify these signals, ranging from statistical feature-based approaches, such as empirical distribution, higher-order cumulants, and cyclostationarity, to modern deep learning (DL)-based models like convolutional neural networks (CNN) and hybrid openset recognition (OSR) architectures. In this paper, we provide a comprehensive review of state-of-the-art blind signal classification algorithms, systematically categorizing existing methods and analyzing their underlying principles. Furthermore, we compare these approaches in terms of computational complexity, robustness to channel impairments, and generalization to unseen signal classes. Based on our analysis, we discuss open challenges and future research directions for non-cooperative signal classification.

키워드