A Stable Modulation Classification Method in Dynamic Environments Using Improved Genetic Algorithm

Abstract

With the development of software radio and cognitive radio technology, much work so far has focused on automatic classification of communication signals, but it still cannot meet the requirements in dynamic environments. Starting from the robustness of feature set, a communication signal modulation classification method with dynamic signal noise ratio(SNR) adaptability is proposed. Firstly, the four categories of features of the signal, such as the instantaneous features of the signal, the high-order cumulant features, the fractal theory features and the entropy features, are extracted to form the original feature set. Besides, the anti-noise evaluation function is constructed by analyzing the variation law of each feature under different SNR. Then, we get a robust feature set through this evaluation function. But there are still a lot of problems such as poor classification ability and information redundancy, thus, the second reduction based on our improved genetic algorithm(GA) is proposed. The experimental results show that the new algorithm improves the classification performance of feature subsets at the expense of time cost.