Red deer optimization for automatic modulation classification using hybrid extreme learning machine with bagging classifier

Abstract

Automatic modulation classification (AMC) became an important process in various communication systems including commercial, telecommunication and military applications. Further, the accuracy of AMC impacts the performance of these applications. Various machine learning approaches were developed to improve the performance of AMC. However, they failed to classify the different modulation schemes, which needs to satisfy all the spectrum requirements under multipath fading environment. Further, the conventional methods suffer with computational complexity in training to satisfy the real-time operational requirements. So, this paper focuses on implementation of extreme learning machine (ELM) for reduction of training complexities and improves the classification performance. Initially, deep leaning convolutional neural network (DLCNN) model is introduced for extracting the interdependent modulation features based on different modulation types. Then, red deer optimization algorithm (RDOA) is introduced for selecting the best features from DLCNN extracted features. Further, the hybrid ELM with bagging classifier (HELM-BC) is used to classify the various modulation types, i.e., families. The simulation results show that the performance of the proposed AMC system using RDOA-based HELM-BC approach is superior to the conventional AMC systems.