A Novel Encoding Deep Learning Algorithm of Radio Signal Understanding

With the social development, more and more new electronic equipment are used and there are more complicated radio signals around us. For the electromagnetic security, it is necessary to understand the radio signal and find the potential risk or abnormal behavior. With deep learning methods are used in many practical products, especially in compute vision field. So, in this paper, the radio signal is encoded into the Gramian Angular Summation Fields (GASF) image by two phase encoding, then using the simple Convolutional Neural Network(CNN)method to process the radio signal. The method encodes the in-phase and the quadrature (IQ) components into GASF respectively. Then we use the combine Two Phase GASF (TPGASF)images to recognize the radio signal modulation through the CNN method. In this paper, experiment is set up with the radio signal modulation which is a typical scene of radio signal understanding. The experiment results show that the more accuracy is got through the TPGASF algorithm than other methods.