Interference classification and identification of TDCS based on improved convolutional neural network

Abstract A classification and recognition algorithm based on short-time Fourier transform and convolutional neural network (STFT-CNN) is proposed to solve the common interference signal classification and recognition problem in transform domain communication systems. In this algorithm, the time-spectrum diagram of interference signals obtained by short-time Fourier transform is input into the vggnet-16 network model improved according to STFT characteristics for feature learning and training, and the classification and recognition of signals are completed. Simulation results show that the proposed algorithm for comprehensive recognition rate reached 97.7%, 6 kinds of jamming signal in low SNR circumstance still can reach more than 93% recognition rate, compared with the traditional algorithm, this method not only improves the classification recognition rate of single interference, but also improves the recognition of mixed interference ability, has the ability to resist low signal-to-noise ratio, makes the transform domain communication system can choose transform domain for anti-interference, provides theoretical basis and support for the application of convolutional neural network in anti-interference of communication system in transform domain.