基于自动机器学习的航天测控网频谱识别方法

In order to realize the reasonable allocation of spectrum resources in the space TT&C network, the first step is to have a high accuracy rate for its spectrum identification. In this paper, the power spectral density sampling data of Loo model and Corazza model, which are typical channel models of space TT&C system, are used as input data sources. Using automatic machine learning technology, different neural architecture search algorithms are described, and the input data are trained by genetic algorithm, gradient descent and Bayesian optimization respectively, and the accuracy of spectrum recognition is obtained and compared. The experimental results show that the three algorithms have their own advantages and disadvantages, and the Bayesian optimization method can obtain higher spectral recognition accuracy with fewer iterations.