Radio Frequency Machine Learning (RFML), wireless signal machine learning technology, extensively extracts signal features through neural network, so as to achieve the identification of drone signals, IDs and other functions. In the recognition and defense of low-altitude drone, its neural network does not depend on the special design of the transmitted signal, and can extract and use the signal features of a longer range, and the machine learning time is shorter, can quickly identify different types of low-altitude drone, so as to realize the signal learning of unknown UAV and the rapid expansion and improvement of the model library. Compared with traditional UAV detection and recognition technology, RFML has the following advantages:

1. Fast support for unknown drones

For the support of DIY, traversal and other new drones, only a certain amount of UAV signals need to be obtained, and then the support of the new drones can be achieved through the training of neural network. Compared with the traditional manual extraction of signal features, the training time of neural network has significant advantages such as short time and less manual participation.

2. Wider range of radio applications

Traditional UAV detection techniques mostly rely on the special design of the transmitted signal, and it is difficult to take advantage of other unique characteristics of the signal (such as CFO). However, these features can be extracted and exploited by RFML using neural networks. In other words, RFML can take advantage of a much wider range of signal characteristics than traditional technologies, so RFML can be applied to a much wider range of applications than traditional technologies.