Radio signal intelligent recognition algorithm based on deep learning

Most of the existing radio signal classification and recognition methods rely on artificial feature extraction or neural networks to learn features of radio signals. The main problems are as follows: first, it is difficult to manually extract features, it is greatly affected by interference signals, and the classifier cannot handle the simultaneous occurrence of multiple radiation sources and signal spectrum aliasing; Second, in the use of deep learning, in order to fully extract features, the network structure is generally very deep, resulting in a relatively long training time for data sets; Third, in the actual application scenario, if there is a new model that has not been collected, it cannot be identified, and it is difficult to achieve fast online training. In this paper, an algorithm combining YOLO prediction boundary box and feature extraction is proposed, which fully considers the recognition effectiveness and real-time by virtue of deep learning, and can also realize rapid training in the face of unknown radiation sources, which is of great application value.