A Two-Stage Hybrid Feature Selection Method of Communication Signals Based on Combined Scoring and Improved Binary Whale Optimization Algorithm

Abstract

Communication signals have many modulation types in the current complex electromagnetic environment, so it is more and more difficult to recognize them accurately. Extracting features from different dimensions can describe the essence of the signals in different aspects and better distinguish various signal modulation formats. However, there may be features with poor anti-noise performance, irrelevant features and redundant features in the original high-dimensional feature set, which not only reduce the recognition probability of the signals but also make the feature selection space grow exponentially. Aiming at the problem of multi-dimensional feature selection of communication signals, this paper proposes a two-stage hybrid feature selection method based on combined scoring and improved binary whale optimization algorithm. In the first stage, the method uses the combined algorithm to initially filter the extracted signal features, and in the second stage, the improved binary whale optimization algorithm is used for further optimization. The results show that, compared with the original feature set, this method can increase the average recognition probability of the signals by up to 17%, and the feature reduction rate up to 81%. Besides, we prove that the proposed algorithm has certain generalization ability for different signal-to-noise ratio (SNR) and classifiers.