Publication: Intelligent automatic single and multi-carrier modulation classification methods
Loading...
Date
2023-10-01
Authors
Dhamyaa Husam Al-Nuaimi
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Automatic modulation classification (AMC) is utilized for modulation type classification in many wireless communication applications, such as electronic surveillance systems and warfare. However, existing AMC methods have three main limitations, namely (i) inefficient feature extraction, (ii) poor classification accuracy for the low signal-to-noise ratio (SNR) rates, and (iii) high computational time. To overcome these issues, two new AMC methods have been proposed, namely AMC using a feature clustering‑based two‑lane capsule network (AMC2N) and an intelligent pyramid model for automatic multi-carrier modulation classification (AMC2-Pyramid). The proposed AMC2N is designed to classify single-carrier modulations, which consists of three phases. In the first phase, tri-level preprocessing is introduced to improve signal quality. In the second phase, two-lane capsule network (TL-CapsNet) is applied to extract the signal’s real and imaginary parts, before neutrophic c-means (NCM) is used to cluster the extracted features. The modulation schemes are then classified in the last phase using SoftMax layer. On the other hand, AMC2-Pyramid is proposed to classify multi-carrier modulation. At the beginning, bi-folding method is used to enhance signal quality. A new pyramidal approach is introduced to extract signal’s features, which is then clustered using human mental search (HMS). Modulation classification is finally implemented using a combination of multi-distance-based nearest-centroid classifier (MdNC2) and improved Q-learning (IQL). For the single-carrier modulation, the proposed AMC2N produces the highest average classification accuracy of 94.73%, compared to 52.98% to 65.67% for other AMC methods. The computational time of the proposed AMC2N is 190.83 ms, which is lower than that of the other AMC methods. For the multi-carrier modulation, the proposed AMC2-Pyramid produces 95.43% accuracy, compared to 58.30% to 86.50% for other AMC methods. The computational time of AMC2-Pyramid is the lowest at 139 ms. The results demonstrate the high accuracy of the proposed methods, AMC2N and AMC2-Pyramid, in classifying modulation types for single- and multi-carrier signals.