Signal Classification in Real-time Based on SDR using Convolutional Neural Network

In the modern complex battlefield environment, traditional conventional parameter measurement, fixed template matching and other methods are difficult to accurately identify unknown target signals and determine threats, and could not implement effective electronic attack. To solve the above problems, this paper designs a software radio system based on (Software-Defined Radio)SDR deep learning real-time signal classification. The system adopts the optimized convolutional neural network and uses training samples with different SNR(Signal-Noise Ratio) to enhance the robustness of the recognition network in various SNR environments. Finally, over the air test is realized by Pluto SDR platform, the accuracy of signal classification can reach at 96.83%. The system can complete the classification and recognition of typical radio signal modulation patterns in the battlefield. It has the advantages of high recognition accuracy and flexible deployment. It provides a new method and means to solve the difficulties of communication target reconnaissance and jamming.