Performance evaluation of the correntropy coefficient in automatic modulation classification

• Published in: Expert systems with applications Vol. 42; no. 1; pp. 1 - 8
• Main Authors: Fontes, Aluisio I.R., de M. Martins, Allan, Silveira, Luiz F.Q., Principe, J.C.
• Format: Journal Article
• Language: English
• Published: Amsterdam Elsevier Ltd 01.01.2015; Elsevier
• Subjects: Applied sciences; Automatic modulation classification; Automation; Channels; Classification; Coefficients; Computer science; control theory; systems; Computer systems and distributed systems. User interface; Correntropy coefficient; Detection, estimation, filtering, equalization, prediction; Equipments and installations; Exact sciences and technology; Information theory; Information, signal and communications theory; Mathematical models; Mobile radiocommunication systems; Modulation; Modulation, demodulation; Radiocommunications; Signal and communications theory; Signal, noise; Software; Telecommunications; Telecommunications and information theory; Wireless communication; Correntropy coefficient; Automatic modulation classification; Information theory; Performance evaluation; Automatic classification; Scalability; Similarity; Statistical moment; Adaptive system; High precision; Entropy; Pattern recognition; Kernel method; Information measure; Measure theory; Surveillance; Wireless network; Metric; Automatic recognition; Adaptive modulation; Signal detection; Monitoring; Software radio; Signal to noise ratio
• ISSN: 0957-4174; 1873-6793
• DOI: 10.1016/j.eswa.2014.07.023

•We propose the direct use of the correntropy for modulation format recognition.•We investigate the influence of kernel size on the performance of the classifier.•The absence of a pre-processing module in the classifier reduces the complexity.•The method was shown to be highly efficient and scalable for wireless systems. Automatic modulation classification (AMC) techniques have applications in a variety of wireless communication scenarios, such as adaptive systems, cognitive radio, and surveillance systems. However, a common requirement to most of the AMC techniques proposed in the literature is the use of signal preprocessing modules, which can increase the computational cost and decrease the scalability of the AMC strategy. This work proposes the direct use of a similarity measure based on information theory for the automatic recognition of digital modulations, which is known as correntropy coefficient. The performance of correntropy in AMC applied to channels subject to additive white Gaussian noise (AWGN) is evaluated. Specifically, the influence of the kernel size on the classifier performance is analyzed, since it is the only free parameter in correntropy. Besides, a relationship between its respective value and the signal-to-noise ratio (SNR) of the channel is also proposed. Considering the investigated modulation techniques, numerical results obtained by simulation demonstrate that there are high accuracy rates in classification, even at low SNR values. By using correntropy, AMC task becomes simpler and more efficient.