Advances and Emerging Challenges of Digital Modulation Recognition for Wireless Signals Using Artificial Intelligence Techniques

Abstract

This paper discusses a statistical review of various artificial intelligence techniques used in wireless communication for the implementation of automatic modulation recognition of digital modulated signals. Automatic recognition of modulation types from received signals in communication systems reduces complexity and cost of circuit design of the demodulators used at receiver. It helps to design an adaptive universal communication receiver in a noisy medium, where the behavior of a modulated signal keeps on changing continuously with noise. For the effective usage of RF (radio frequency) spectrum, determination of modulation types in spectrum sensing and software defined radios (SDRs) is considered to be crucial. Demodulation and recognition of signal types using hybrid machine learning algorithms at receiving end plays very significant role in various applications of cognitive radios, satellite communication systems, and military intelligence and in underwater surveillance systems too. Implementation methodologies of these applications for automatic recognition of modulation types are discussed using various AI techniques. Various spectral and statistical parameters like time-frequency features, higher order cumulants, constellation images and cyclic frequency domain features extracted from received signals for classifying modulation types are summarized. Different algorithms developed to extract features, classification algorithms and the validation metrics used in the models are outlined. Methodologies used to distinguish the type of digital modulated signals like ASK, PSK, FSK and QAM signals with its inferences and analysis from datasets are discussed.