Abstract
Cyclostationary analysis has several applications in communications, e.g., spectral sensing, signal parameter estimation, and modulation classification. Most of them consider the additive white Gaussian noise (AWGN) channel model, although wireless communication systems may also be subject to non-Gaussian interference and impulsive noise. In this context, the communication channel can be better modeled by heavy-tailed distributions, such as the non-Gaussian alpha-stable one. Some applications of the cyclostationary approach based on the spatial sign cyclic correlation function (SSCCF), fractional lower-order cyclic autocorrelation function (FLOCAF), and cyclic correntropy function (CCF) demonstrate that these are promising solutions for the analysis of signals in the presence of impulsive non-Gaussian noise. However, the investigation of functions above applied to digital modulation recognition in impulsive environments, and the comparison among them are topics that did not adequately explore yet. This work demonstrates that SSCCF is a particular case of the FLOCAF. Besides, a detailed analysis of the use of the FLOCAF and CCF is presented to obtain cyclostationary descriptors for the recognition of digital modulations BPSK, QPSK, 8-QAM, 16-QAM, and 32-QAM. Automatic modulation classification (AMC) architectures, based on the functions mentioned above, are also proposed. Besides, another contribution showed is that both the FLOCAF and CCF allow the symbol rate parameter estimation. The performances of AMC architectures were evaluated in the scenario with modulated signals contaminated with additive non-Gaussian alpha-stable noise. The results demonstrate that both architectures can classify signals in different contamination scenarios. However, the architecture based on the CCF is more efficient than the FLOCAF-based one.
Highlights
The purpose of automatic modulation classification (AMC) is to identify the unknown modulation format of the received noisy signal, in a short period with a hit-rate as high as possible [1], [2]
The hit-rate of this modulation growing slower, which indicates that 16-QAM is more sensitive to noise contamination than other analyzed modulations. These results indicate that, for the classification criteria adopted in the proposed architecture, the cyclic signatures of BPSK, 8-QAM, and, 32-QAM are the most robust to impulsive noise contamination, once that, even for some negative geometric signal-tonoise ratio (GSNR) values, the classification hit-rates for these modulations are elevated
In the cyclostationary signatures provided by the fractional lower-order cyclic autocorrelation function (FLOCAF) and cyclic correntropy function (CCF), the component associated with the symbol rate of the modulated signal is the most robust to impulsive noise contamination, as demonstrated in Fig. 19 and Fig. 20
Summary
The purpose of automatic modulation classification (AMC) is to identify the unknown modulation format of the received noisy signal, in a short period with a hit-rate as high as possible [1], [2]. Cyclostationary analysis allows the extraction of cyclic spectral features from communication signals, known as cyclostationary signatures, and can be efficiently used in Gaussian environments for spectral sensing [22], [23], automatic modulation recognition [23]–[25], and estimation of signal parameters [20], [23], [26]. A detailed analysis about the cyclic spectrum of BPSK, QPSK, 8-QAM, 16-QAM, and 32-QAM signals obtained by the FLOCAF and CCF is provided, evidencing that such functions can extract singular descriptors capable of distinguishing the modulations above, even in environments contaminated by non-Gaussian alpha-stable noise;.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
