Classification of multi-carrier digital modulation signals using NCM clustering based feature-weighting method

Abstract

This work presents a novel digital modulation signal classification model by combining Neutrosophic c-means (NCM) based feature weighting (NCMBFW) and classifier algorithms. As the digital modulation signal, the multi-carrier amplitude shift keying (MC-ASK), frequency shift keying (MC-FSK), and phase shift keying (MC-PSK) modulation types are employed. In the first step, the feature extraction process has been conducted from the raw digital modulation signals and thereby extracted time, frequency, and time-frequency domain features from the multi-carrier ASK, FSK, and PSK signals. After that, these features have been weighted by using NCMBFW. Finally, classifier algorithms including Linear Discriminant Analysis (LDA), Support Vector Machine (SVM), k-nearest neighbor (k-NN), AdaBoostM1, and Random Forest, have been used to determine the types of digital modulation signals automatically. Many metrics are used to evaluate the performance in the experiments. The proposed method in the classification of MC digital modulation signals is the first work with respect to the classification of MC modulation signals in the literature. For worst case (in 5 dB), while the obtained f-measure values are 0.842, 0.848, 0.863, 0.842, and 0.894 using LDA, SVM, k-NN, AdaBoostM1, and Random Forest classifiers without NCMBFW, respectively, while the f-measure values by combining NCMBFW with classifier algorithms are 0.983, 0.976, 0.992, 0.988, and 0.991, respectively. The experimental results show that the proposed NCMBFW can be considered as a promising tool to improve the classification performance of digital multi-carrier modulation signals.