Arquiteturas de classificação automática de modulações em ambientes impulsivos baseadas em características cicloestacionárias

Abstract

The rapid growth of wireless communications systems’ applications drives new strategies for efficient exploration of the frequency spectrum, such as cognitive radios. The cognitive radio is an intelligent communications system capable of autonomously adapting to the communication channel by reconfiguring its operating parameters. An essential property of cognitive radios is automatically recognizing the type of modulation employed in a Radio Frequency (RF) signal, thus enabling interoperability between systems, improving spectral efficiency, or even enabling electronic surveillance. This attribute is known as automatic modulation classification (AMC). Techniques based on the detection of patterns obtained from the analysis of cyclostationary characteristics of the second-order compose the state-of-the-art of AMC applications. Although very widespread, these techniques cannot recognize some types of digital high-order modulations, such as M-ary quadrature amplitude modulation (MQAM) and M-ary phase-shift keying (MPSK) modulations. On the other hand, the higher-order cyclostationary analysis techniques used to extract particular descriptors of these modulations have a very high computational cost. Besides, they are only suitable for communication environments with additive white gaussian noise (AWGN). Although the AWGN model is widely used to characterize wireless communication channels, several practical scenarios are better modeled by non-gaussian distributions, such as high frequency (HF) communication, whose environment presents substantial contamination by impulsive noise. In this scenario, two emerging cyclostationary analysis functions, the fractional lower-order cyclic autocorrelation function (FLOCAF) and the cyclic correntropy function (CCF), were efficiently employed in impulsive communications environments. However, these functions were evaluated on the impulsive channel’s spectral sensing, a less complicated problem to the automatic modulation classification. Since there is no satisfactory solution in the literature for the automatic classification of high-order modulations in channels with impulsive noise, this work proposes automatic modulation classification architectures based on FLOCAF and CCF, combined with the decision tree classification and logistic regression techniques. The architectures were developed to recognize digital modulations Binary PSK (BPSK), Quadrature PSK (QPSK), 8-QAM, 16-QAM, and 32-QAM, and evaluated in different contexts of contamination by additive impulsive noise, modeled by alpha-stable distribution. The results showed that all architectures could operate in impulsive environments, with CCF-based architectures being the most efficient.