A new method to solve the problem of facing less learning samples in signal modulation recognition
Abstract
Abstract In machine learning method, the number of training samples is an exceedingly important factor determining the learning system’s robustness. In our previous researches (Liu et al., J. Syst. Eng. Electron. 27.2:333–342, 2016; Liu et al., IET Commun. 11.7:1000–1007, 2017), the extreme learning machines (ELMs) have proven to be an effective and time-saving learning method for pattern classification and the signal modulation recognition. ELMs are utilized to supervised learning issues principally on signal modulation recognition. In this thesis, ELMs are extended for semi-supervised tasks that are based on the manifold regularization, therefore greatly enlarging ELMs’ applicability. This article evolves countermeasures to the less training samples which mitigate the modulation recognition efficacy and demonstrates the robustness of semi-supervised learning for signal classification in AWGN and Rayleigh-fading channels.